NightShadowPT wrote on 2023-08-10, 05:47:

Thanks, that clarifies it. […]
Show full quote

PC Hoarder Patrol wrote on 2023-08-09, 10:22:

Presumably this...

50M Weitek.jpg

Thanks, that clarifies it.

Now for the million dollar question.
I understand why you would want a 387 for a 386 processor... I know their uses are limited, but they exist, but if the 486DX already included an FPU, why would anyone need a 387 for?

https://www.cpu-world.com/CPUs/4167/index.html
Seems like the Weitek 4167 was meant to be paired with early 486 CPUs.
But the Weitek instruction set was not x87 compatible so programs needed to be specifically compiled to support the Weitek instructions.
Very interesting. I thought FPUs ended after the 386 era.

Well, it's done... but unfortunately the mod has failed.

I changed the oscillator on the board, but ended up with graphical corruption (but I could make out the computer was doing the memory test, and passed it).

Unfortunately, this seems to mess up the system somehow, so maybe just changing the oscillator and the processor is not enough to transform a 50M board into an 66M 🙁

Back to my procurement exercise... if anyone has a 66M board for sale, please let me know - I would still like to upgrade my 486 DX2/66 with a board with L2 cache.

🙁

Something doesn’t like being overclocked. Very strange for 1991, not even able to click 33mhz

When you move back to 50mhz everything is fine?

There used to be strange oscillator speeds 55mhz 60mhz etc, you could try to see if the board will overclock at all.

There are also 75mhz and 100mhz overdrives that will work on a 25mhz FSB that might be drop in replacements so you don’t need to source a new card

NightShadowPT wrote on 2023-08-14, 18:40:

Well, it's done... but unfortunately the mod has failed. […]
Show full quote

Well, it's done... but unfortunately the mod has failed.

I changed the oscillator on the board, but ended up with graphical corruption (but I could make out the computer was doing the memory test, and passed it).

Unfortunately, this seems to mess up the system somehow, so maybe just changing the oscillator and the processor is not enough to transform a 50M board into an 66M 🙁

Back to my procurement exercise... if anyone has a 66M board for sale, please let me know - I would still like to upgrade my 486 DX2/66 with a board with L2 cache.

🙁

Well that's an unfortunate conclusion. ☹️

rmay635703 wrote on 2023-08-14, 19:43:

Something doesn’t like being overclocked. Very strange for 1991, not even able to click 33mhz […]
Show full quote

Something doesn’t like being overclocked. Very strange for 1991, not even able to click 33mhz

When you move back to 50mhz everything is fine?

There used to be strange oscillator speeds 55mhz 60mhz etc, you could try to see if the board will overclock at all.

There are also 75mhz and 100mhz overdrives that will work on a 25mhz FSB that might be drop in replacements so you don’t need to source a new card

If I see it correctly, the SRAM data chips are 25ns. And only a single bank. So maybe the l2 cache is the problem. Is there a possibility to disable it to test if this is the culprit?

glitchy cache would crash the system, while this distorted picture suggests VGA not liking faster bus somehow.

Where are you located NightShadowPT?
I have some spare 33Mhz boards and I might send you one from germany?
It willjust take some time cause I am on holidays and business travel for the next two weeks.

8kb AMD DX4 - 100mhz chips exist.

And of course 5x86 chips and Intel dx4 overdrives

Ran under clocked these chips will sometimes last years at 5 volts (sometimes not). There are overdrive units that fix the voltage problem

Either should be compatible and won’t require overclocking

eisapc wrote on 2023-08-15, 08:30:

Where are you located NightShadowPT?
I have some spare 33Mhz boards and I might send you one from germany?

I'm living in the UAE, but I have a forwarding address I can use in Germany.

To be clear, I currently have a working board running a 486 DX2/66, but it is one of the boards without L2 cache (part number 129127-001).

This "adventure" was an attempt to modify a 486 DX2/50 board with L2 Cache( part number 129128-001) to run with my 486DX2/66. It unfortunately failed 🙁

To my (admittedly flawed) knowledge there are only two processor boards that support 486 processors and have L2 cache. They have part numbers 129128-001 (for the DX2/50) and 137663-001 (for the DX2/66).

If you have a board with the part number 137663-001, I would be interested in it. Also, if you are aware of another board that has L2 cache that I may be unaware of, happy to discuss.

Thanks in advance and please let me know.

rasz_pl wrote on 2023-08-15, 04:16:

glitchy cache would crash the system, while this distorted picture suggests VGA not liking faster bus somehow.

I would argue the issue does not seem to be with the graphics card running at the higher FSB (I have another card with a DX2/66 that works flawlessly with that same VGA card).

CoffeeOne wrote on 2023-08-14, 21:20:

If I see it correctly, the SRAM data chips are 25ns. And only a single bank. So maybe the l2 cache is the problem. Is there a possibility to disable it to test if this is the culprit?

That may well be the case, but the whole purpose of this endeavor was for me to get a board with L2 cache, so even if this is possible, it would defeat the purpose.

NightShadowPT wrote on 2023-08-17, 06:24:

rasz_pl wrote on 2023-08-15, 04:16:

glitchy cache would crash the system, while this distorted picture suggests VGA not liking faster bus somehow.

I would argue the issue does not seem to be with the graphics card running at the higher FSB (I have another card with a DX2/66 that works flawlessly with that same VGA card).

I would argue that the issue might very well be with the graphics card, and the higher FSB is indirectly to blame.
The (E)ISA bus speed is 8.33MHz which is 50/6 (or 25/3)
now you up the bus to 33 yet the divider is still the same and thus the (E)ISA bus now runs at 11MHz.
It might not seem much but it's still a ~40% increase.

The other system you tested it on, will likely have another divider, like 4, so 33/4 = 8.25 .

Now I hear you coming, "Where do I change that bus divider?" and I can only answer that I do not know but maybe someone else has a clue.

edit: I could also be completely wrong and the (E)ISA bus runs at the same speed no matter the FSB speed, though at the moment I doubt it.

weedeewee wrote on 2023-08-17, 15:07:

I would argue that the issue might very well be with the graphics card, and the higher FSB is indirectly to blame. The (E)ISA bu […]
Show full quote

NightShadowPT wrote on 2023-08-17, 06:24:

rasz_pl wrote on 2023-08-15, 04:16:

glitchy cache would crash the system, while this distorted picture suggests VGA not liking faster bus somehow.

I would argue the issue does not seem to be with the graphics card running at the higher FSB (I have another card with a DX2/66 that works flawlessly with that same VGA card).

I would argue that the issue might very well be with the graphics card, and the higher FSB is indirectly to blame.
The (E)ISA bus speed is 8.33MHz which is 50/6 (or 25/3)
now you up the bus to 33 yet the divider is still the same and thus the (E)ISA bus now runs at 11MHz.
It might not seem much but it's still a ~40% increase.

The other system you tested it on, will likely have another divider, like 4, so 33/4 = 8.25 .

Now I hear you coming, "Where do I change that bus divider?" and I can only answer that I do not know but maybe someone else has a clue.

edit: I could also be completely wrong and the (E)ISA bus runs at the same speed no matter the FSB speed, though at the moment I doubt it.

My apologies, but I am not sure I follow your rationale.

As I mentioned, I have another processor card that works @33Mhz and supports the same 486DX2/66, and the VGA card works without any problems.

Since both cards operate at 33Mhz, I assume we can rule out the FSB being a problem (?) Am I missing something?

NightShadowPT wrote on 2023-08-17, 16:29:

My apologies, but I am not sure I follow your rationale. […]
Show full quote

weedeewee wrote on 2023-08-17, 15:07:

I would argue that the issue might very well be with the graphics card, and the higher FSB is indirectly to blame. The (E)ISA bu […]
Show full quote

NightShadowPT wrote on 2023-08-17, 06:24:

I would argue the issue does not seem to be with the graphics card running at the higher FSB (I have another card with a DX2/66 that works flawlessly with that same VGA card).

I would argue that the issue might very well be with the graphics card, and the higher FSB is indirectly to blame.
The (E)ISA bus speed is 8.33MHz which is 50/6 (or 25/3)
now you up the bus to 33 yet the divider is still the same and thus the (E)ISA bus now runs at 11MHz.
It might not seem much but it's still a ~40% increase.

The other system you tested it on, will likely have another divider, like 4, so 33/4 = 8.25 .

Now I hear you coming, "Where do I change that bus divider?" and I can only answer that I do not know but maybe someone else has a clue.

edit: I could also be completely wrong and the (E)ISA bus runs at the same speed no matter the FSB speed, though at the moment I doubt it.

My apologies, but I am not sure I follow your rationale.

As I mentioned, I have another processor card that works @33Mhz and supports the same 486DX2/66, and the VGA card works without any problems.

Since both cards operate at 33Mhz, I assume we can rule out the FSB being a problem (?) Am I missing something?

Should I ask if you read the whole comment I made or only the first line, or just wait, since it seems to me like you only read the first line and ignored the rest of my previous comment.
edit: I could try and rephrase it, and expand it some more, though the gist would remain the same.

This 50 MHz (25MHz) board which you put another crystal on, and is now running at 66 MHz (33Mhz) likely has a bus divider of 6 (3).
This turns your (E)ISA bus clock from the standard 8.33MHz into a clock of 11MHz

The board with the original 66Mhz (33MHz) CPU likely has a bus divider of 8 (4)
This turns your (E)ISA bus clock into an 8.25MHz clock.

I'm sorry. I don't know how else I can explain it. This should be clear & simple enough.

weedeewee wrote on 2023-08-17, 16:48:

This 50 MHz (25MHz) board which you put another crystal on, and is now running at 66 MHz (33Mhz) likely has a bus divider of 6 ( […]
Show full quote

This 50 MHz (25MHz) board which you put another crystal on, and is now running at 66 MHz (33Mhz) likely has a bus divider of 6 (3).
This turns your (E)ISA bus clock from the standard 8.33MHz into a clock of 11MHz

The board with the original 66Mhz (33MHz) CPU likely has a bus divider of 8 (4)
This turns your (E)ISA bus clock into an 8.25MHz clock.

I'm sorry. I don't know how else I can explain it. This should be clear & simple enough.

It's clear now. Thanks for rephrasing it.

NightShadowPT wrote on 2023-08-17, 20:23:

weedeewee wrote on 2023-08-17, 16:48:

This 50 MHz (25MHz) board which you put another crystal on, and is now running at 66 MHz (33Mhz) likely has a bus divider of 6 ( […]
Show full quote

This 50 MHz (25MHz) board which you put another crystal on, and is now running at 66 MHz (33Mhz) likely has a bus divider of 6 (3).
This turns your (E)ISA bus clock from the standard 8.33MHz into a clock of 11MHz

The board with the original 66Mhz (33MHz) CPU likely has a bus divider of 8 (4)
This turns your (E)ISA bus clock into an 8.25MHz clock.

I'm sorry. I don't know how else I can explain it. This should be clear & simple enough.

It's clear now. Thanks for rephrasing it.

Another question: Did you run the ECU after changing from 25MHz to 33MHz? EDIT: Maybe hard when you don't see something on the screen. Bad idea.

I do not believe the EISA clock is derived from CPU clock on these deskpro/m systems

what clocks exactly are sourced from the cpu carrier? and can OP use another graphic card? for example simple ISA one?

NightShadowPT wrote on 2023-08-17, 06:24:

CoffeeOne wrote on 2023-08-14, 21:20:

If I see it correctly, the SRAM data chips are 25ns. And only a single bank. So maybe the l2 cache is the problem. Is there a possibility to disable it to test if this is the culprit?

That may well be the case, but the whole purpose of this endeavor was for me to get a board with L2 cache, so even if this is possible, it would defeat the purpose.

It would defeat the purpose, yes, but it would bring some valuable information to the table, like it may be an EISA divider issue ...

Try with another VGA card, an ISA one should be usable.