i explained the 3 oscillators in previous post.

ISA bus was my first guess, but I was expecting a 16MHz crystal. Any idea why it needs to be divided by 4?

pshipkov wrote on 2022-06-18, 03:28:

i explained the 3 oscillators in previous post.

link?

Strange I thought I posted an extract from a similar motherboard with the same chipset, essentially it is a manual way of doing the BUSCLK instead of being done in the BIOS, according to the motherboard notes:

Pins 1 and 2: Bus speed select BUSCLK/2 default for 16MHz CPU
Pins 2 and 3: Bus speed select BUSCLK/4 default fro 20 or 25MHz CPU

But not sure if it really has that effect to tell the truth, or if the notes are that acurate.

J1 - In position 1-2 FPU uses the main oscillator. In position 2-3 FPU uses the oscillator next to it. If position 2-3 is selected and that oscillator is not present - no FPU. Oscillator's frequency cannot be higher than the main (first) oscillator.
J2 - No visible change in perf or other parameters of the system.
J9 - Switches between the first and second oscillator (the one on its right hand side) - a bit odd, but i can see how this can be a product feature.

pshipkov wrote on 2022-06-18, 07:21:

J1 - In position 1-2 FPU uses the main oscillator. In position 2-3 FPU uses the oscillator next to it. If position 2-3 is selected and that oscillator is not present - no FPU. Oscillator's frequency cannot be higher than the main (first) oscillator.
J2 - No visible change in perf or other parameters of the system.
J9 - Switches between the first and second oscillator (the one on its right hand side) - a bit odd, but i can see how this can be a product feature.

Speaking of the 80287..
I vaguely remember that it is possible to run with its own clock.
All it needs is a little interposer card and a bog standard 5v TTL crystal oscillator.
Most pins are 1:1, with the exception of the clock pin, and another pin that disables the 2:3 clock divider. Speaking under correction, though.
The 287 will now run at the speed of the crystal oscillator.
That's good to know if the 286 mainboard is limited in that respect.
This way, you can use any 287 chip you want. You can even use an ancient 8 MHz 287 with a 25MHz 286.
Just keep in mind that the intel i80287 is different to the 287XL and the compatibles.
Merely the ancient intel 80287 is a real 287 class FPU.
The others have a more modern 387 core with an 287 pinout.
They also run at 1:1 clock rate essential (3:2 if you will), so a 286 CPU @10 MHz needs a 10MHz 287XL, too (unless run asynchronous), rather than an 8MHz i80287.
There's some difference in accuracy, also.:

"The 8087 and 80287's FPTAN and FPATAN instructions are limited to an argument in the range ±π/4 (±45°), and the 8087 and 80287 have no direct instructions for the SIN and COS functions."

Source: Wikipedia

That's why very old software is perhaps in better hand if it's run with an 8087 software emulator than on a 287XL.
The software emulator is Real-Mode only, though. Because 8087 didn't support Protected-Mode, yet. Unlike the 80287, which works fine with either Protected-Mode (16/32-Bit).
So it's of no use in Windows 3.1x in its "Standard-Mode" - however, 3.x ships with its own emulator/virtualizer, win87em, anyway.

I am not sure it is possible to run 287 fpu completelly async to the cpu/base frequency.
At least i was not able to achieve that so far.
There are few boards that offer 1:1 or 2:3 multiplier controls by effectively pulling up or down voltage on one of the fpu legs.
I am not completelly sure if each and every 287 fpu model can do that, but late cyrix, iit, ulsi and intel models can do it. Also, as you said theg zre 387 cores/backports than native 287 silicon.
Tried also couple of ugly contraptions with alternati e freq but that resulted in not recognized fpu during post.
I admit that these were not scientific experiments at zll, more like me rushedly trying something and loosing interest quickly.

Is there a proven fact that 287 fpu can run completelly async to cpu ?

pshipkov wrote on 2022-06-19, 00:43:

Is there a proven fact that 287 fpu can run completelly async to cpu ?

Notarially certified? 😁

pshipkov wrote on 2022-06-19, 00:43:

I am not sure it is possible to run 287 fpu completelly async to the cpu/base frequency. At least i was not able to achieve that […]
Show full quote

I am not sure it is possible to run 287 fpu completelly async to the cpu/base frequency.
At least i was not able to achieve that so far.
There are few boards that offer 1:1 or 2:3 multiplier controls by effectively pulling up or down voltage on one of the fpu legs.
I am not completelly sure if each and every 287 fpu model can do that, but late cyrix, iit, ulsi and intel models can do it. Also, as you said theg zre 387 cores/backports than native 287 silicon.
Tried also couple of ugly contraptions with alternati e freq but that resulted in not recognized fpu during post.
I admit that these were not scientific experiments at zll, more like me rushedly trying something and loosing interest quickly.

Is there a proven fact that 287 fpu can run completelly async to cpu ?

I have a motherboard that runs it off the KBC clock, so yes

^Thank you. ^^

Personally, I've seen a photo of an interposer card in an magazine from the 1980s.
CHIP magazine, I think, so circa 1985 to 1989..
The card roughly was the size of two lighters and
had a crystal oscillator and an 287 in a socket.

Unfortunately, I have absolutely no idea were the issue is now.
Maybe in the attic. I don't have immediate access to it, unfortunately.

Edit: I think it's possible that the card had an additional TTL chip, also.
Some sort of buffer, to clean up the square pulses from the oscillator for stability reasons.
Anyway, I'm not sure. Maybe it's the Mandela effect, not sure.

@Jo22
🤣.
Community certification will be enough i guess.

Did both of you confirmed that when the FPU is ticking at different rate using the auxiliary oscillator the performance varies ?
Asking this because on the motherboard in question here (and at least 2 other boards) i can supply alternative frequency, but somehow perf does not change.
I guess i can easily show this on video with this board for example.
Hope you can follow my description here:
1. Jmper in position that takes frequency from oscillator nearby.
2. If no oscillator - no FPU. This is proof that frequency comes from there.
3. Case 1 - insert 16MHz oscillator - FPU perf is at the same level when using the main oscillator.
4. Case 2 - insert 30MHz oscillator - FPU perf is at the same level when using the main oscillator.
5. Main crystal is at 30MHz. If i insert FPU oscillator above 30MHz - no lights.

As i mentioned above - remember seeing similar thing on other boards as well.
Maybe these alternative frequencies are not really driving the FPU directly, but only used for some sync or who knows what logic and the actual frequency is coming from the main oscillator.
That's why asking if it was verified that perf varies in your cases.

pshipkov wrote on 2022-06-19, 17:43:

@Jo22 LOL. Community certification will be enough i guess. […]
Show full quote

@Jo22
🤣.
Community certification will be enough i guess.

Did both of you confirmed that when the FPU is ticking at different rate using the auxiliary oscillator the performance varies ?
Asking this because on the motherboard in question here (and at least 2 other boards) i can supply alternative frequency, but somehow perf does not change.
I guess i can easily show this on video with this board for example.
Hope you can follow my description here:
1. Jmper in position that takes frequency from oscillator nearby.
2. If no oscillator - no FPU. This is proof that frequency comes from there.
3. Case 1 - insert 16MHz oscillator - FPU perf is at the same level when using the main oscillator.
4. Case 2 - insert 30MHz oscillator - FPU perf is at the same level when using the main oscillator.
5. Main crystal is at 30MHz. If i insert FPU oscillator above 30MHz - no lights.

As i mentioned above - remember seeing similar thing on other boards as well.
Maybe these alternative frequencies are not really driving the FPU directly, but only used for some sync or who knows what logic and the actual frequency is coming from the main oscillator.
That's why asking if it was verified that perf varies in your cases.

I did verify yes, in fact I ended up pulling the FPU clock pin out of circuit on the aforementioned motherboard to over clock it. The performance increase was close to linear with the input clock.

Also NSSI will report the FPU clock with a semi-reasonable degree of accuracy

thanks for the confirmation.

Jo22 wrote on 2022-06-18, 16:39:

Merely the ancient intel 80287 is a real 287 class FPU.
The others have a more modern 387 core with an 287 pinout.

Isn't the NMOS 287 more like an 8087 class FPU with a 286 cpu interface?

Anonymous Coward wrote on 2022-06-19, 23:54:

Jo22 wrote on 2022-06-18, 16:39:

Merely the ancient intel 80287 is a real 287 class FPU.
The others have a more modern 387 core with an 287 pinout.

Isn't the NMOS 287 more like an 8087 class FPU with a 286 cpu interface?

And is an officially IEEE compliant instruction set.

8087 could actually outperform the 287 due to having its own mmu

The last time i compared two very fast machines 8086 and 286 at same clock speed - 8086 just didn't stand a chance.

A lot of interest in these oscillators, not socketed the one by the 287 as I ran out of sockets, so ordered some new ones and may install and run tests in the future.

So the current lay down of the build is this:

1) WG-5000 ET4000 video card, apparently made by Diamond Flower, Inc. 1mb RAM but it is not accepting the addition 1mb to make it 2mb

2) Adaptec AHA-1542CF (for both floppy drives, CD and a SCSI hard drive), going to run this as the main drive and use the IDE for data transfer.

3) 3Com Etherlink III 3C509B-C, did try a slightly more period correct one but had issues with the drivers.

4) Generic ALS100 Plus sound card, FCC ID suggests it is by Procomp Informatics, Ltd. Working because of the amazing UNISOUND - Universal ISA PnP Sound Card Driver for DOS v0.81b

5) Generic 9600 baud modem (to play with ATI commands) configured as COM2

6) Fan Card II I/O Card (for parellel, serial and one IDE for CF card for data transfer)

I am fairly happy with those cards at the moment.

Only issue I am having is since cleaning the SIMM slots, it now does not accept the 4mb SIMM Modules again!
But will accept the 1mb SIMM Modules, so the sockets are not damaged something else must be going on.

So at the moment running on 5mb (1mb dram and 4mb simm).

Things still to do:

1) Install a Mr. BIOS once my new programmer arrives.
2) Investigate what the hell is going on with the SIMM's
3) Maybe, install a socket for a crystal next to the 287 for testing purposes
4) Maybe, Upgrade the video card to 2mb but that needs further investigation as well
5) Maybe, a slight overclock nothing too crazy but need to solve the other issues first.

Software I have been playing with over the weeked:

MS-DOS 5.0 and obviously DOS games.

Windows for workgroups 3.1, just because I have been playing with the NetBEUI protocol with another system.

Happy with this at the time being.

About the oscillators and other side convos here:
Forum conversations were K6/P4 dominated recently so we use your thread for a change. 😀

Like the setup quite a bit.

What you do with the modem ?
Do you have a land line, or some adapter through mobile device gets you online ?

Get these SIMM slots under control already. : )

Will Mr.BIOS give you anything more in addition to the 386 BIOS this mobo comes with ?

You can get this mobo going easily at 30MHz, but most likely without the DIPP RAM chips.

I can recommend the Viking DLE-200B line simulator if you want to play around with modems

simon_e_hall wrote on 2022-06-20, 15:06:

5) Generic 9600 baud modem (to play with ATI commands) configured as COM2

Cool, I also kept some modems.
The internal types do have a fast FiFo directly built-in,
so there's no trouble with using outdated/buggy 8250 or 16450 UART FiFos.

Personally, I've always planned to connect two PCs via internal modems for playing games with my sister.. *sigh*

Anyway, you don't need a real land line.
You can use some old Fritzbox with two telephone ports.
So you can call the other port.

The Fritzbox will also emulate the landlines dial tone sufficiently.
However, it has one limitation. The voltage isn't high enough for classic, electro-mechanical telephone bells.
Electronical telephones with an electric bell (duch as modems) do work fine, though.

Other uses I can think of, are emulated online services of the past.
Things like BTX or Minitel are being emulated by enthusiasts.
It's possible to use some old software like Amaris BTX etc to call such a recreation.

Edit: Another example: CompuServe or AOL.
It's still possible to use the old software on a retro PC in conjunction with similar community projects.

Also, it's possible to built a simple landline at home. My father and me built one some time ago.

Edit: Photos taken and attached. Such a simple landline will create a buzzing, not a ringing sound, however. 😂