Hi Jan,

I have tried adding an ISA slot to a GA-7ZXE rev 2.1 board which has footprint for one, but I have run into problems and I'm wondering if BIOS lacks some functionality that is needed for ISA. So far I have enabled some DMA configs for the BIOS with AMIBCP and added a missing passive component to board (DACK5 was not connected to pin on ISA slot at all, adding a resistor or piece of wire was needed). Now AWE64 gets correctly assigned to IRQ 5 and DMAs 1 and 5. However it's still not functional, Creative's diagnose.exe correctly recognizes the IRQ and DMA settings and is able to play some garbled audio, but CTCM complains about incorrect Vendor ID and Windows gives Code 9 error.

I have been using version FB BIOS downloaded from Gigabyte's site, it and the version modded with AMIBCP are in the attached zip file.

Can you test 8 bit sound/mode/drivers on that card?
I have a hunch that there's a another dead trace in the 16bit area.

myne wrote on 2026-05-28, 11:20:

Can you test 8 bit sound/mode/drivers on that card?
I have a hunch that there's a another dead trace in the 16bit area.

That's where my mind goes to, a bridged or dead line causing data corruption

I have already traced most of the traces in 16 bit area to test points near south bridge, and they seem to be fine. I suppose other DACK/DRQ lines than 5 and IRQs don't matter as they aren't used anyways?

Diagnose.exe's 8bit test was also a bit garbled, but in a regular and not random way. Same as 16bit test actually. However, I tried TIE fighter's install program and I could set it as a Sound Blaster 2.0 and the digital sound test would sound okay-ish. Music test didn't work, not even FM one.

Wait.
Don't you need an xtal too?
Pretty sure they needed a boost over the pci clock.

myne wrote on 2026-05-29, 16:10:

Wait.
Don't you need an xtal too?
Pretty sure they needed a boost over the pci clock.

I think I actually measured ISA clock from the bus and it looked fine. But apparently AWE64 doesn't use it, as there's no contact on that pin.

It probably is divided by four from PCI clock and comes from south bridge. What do you mean with a boost?

I just meant I thought the Southbridge needed its own isa xtal because the pci one was too weak to drive both.

You're on a via chipset. Yes?
Those can run different isa clocks. You might want to find the datasheet, and wpcredit or similar.

About a year or so ago I read a bunch of the datasheets from that era so I shouldn't be too confused.

It's possible the board conversation tools in my sig could help. I doubt there's much difference between an Asus and gigabyte on a via chipset on the isa side.
I can send a few hundred converted Asus boards if there is somewhere to upload. (~200mb)

myne wrote on 2026-05-30, 00:29:

I just meant I thought the Southbridge needed its own isa xtal because the pci one was too weak to drive both. […]
Show full quote

I just meant I thought the Southbridge needed its own isa xtal because the pci one was too weak to drive both.

You're on a via chipset. Yes?
Those can run different isa clocks. You might want to find the datasheet, and wpcredit or similar.

About a year or so ago I read a bunch of the datasheets from that era so I shouldn't be too confused.

It's possible the board conversation tools in my sig could help. I doubt there's much difference between an Asus and gigabyte on a via chipset on the isa side.
I can send a few hundred converted Asus boards if there is somewhere to upload. (~200mb)

A crystal doesn't drive clock directly, but there are various circuits such as probably a PLL and an output buffer before it actually gets to the bus.

It is KT133A chipset, so south bridge is 686B. From the datasheet you posted, it looks like ISA clock really is just generated from PCI clock with a divider, after which there's probably another buffer.

Do you have something like a PicoGUS where you can do data reads over the ISA bus with an expected value?

The nature of the corruption may reveal itself

NeoG_ wrote on 2026-05-30, 11:12:

Do you have something like a PicoGUS where you can do data reads over the ISA bus with an expected value?

The nature of the corruption may reveal itself

Unfortunately I don't. 🙁 I was thinking the same that some sort card that can be used for debugging ISA might be needed to solve this. Although PicoGUS seems to be 8bit so it won't help if the problem is in the 16bit section.

I did buy a POST diagnostics card from Aliexpress, and it'll probably arrive next week. Maybe that'll help with checking whether basic ISA functionality exists?

Mike_ wrote on 2026-05-30, 10:23:

A crystal doesn't drive clock directly, but there are various circuits such as probably a PLL and an output buffer before it act […]
Show full quote

myne wrote on 2026-05-30, 00:29:

I just meant I thought the Southbridge needed its own isa xtal because the pci one was too weak to drive both. […]
Show full quote

I just meant I thought the Southbridge needed its own isa xtal because the pci one was too weak to drive both.

You're on a via chipset. Yes?
Those can run different isa clocks. You might want to find the datasheet, and wpcredit or similar.

About a year or so ago I read a bunch of the datasheets from that era so I shouldn't be too confused.

It's possible the board conversation tools in my sig could help. I doubt there's much difference between an Asus and gigabyte on a via chipset on the isa side.
I can send a few hundred converted Asus boards if there is somewhere to upload. (~200mb)

A crystal doesn't drive clock directly, but there are various circuits such as probably a PLL and an output buffer before it actually gets to the bus.

It is KT133A chipset, so south bridge is 686B. From the datasheet you posted, it looks like ISA clock really is just generated from PCI clock with a divider, after which there's probably another buffer.

The attachment isa_clock.png is no longer available

Yeah, but the pci clock comes from the same clockgen chip that does the cpu and ram. I seem to recall SBs needing a secondary crystal even if the result lines up with pci/x.
In any case, the board I saw a pic of without isa had one next to the chipset so I assume ops does too.

Op take a look in my sig for the bx reference design.
Might give some ideas.

myne wrote on 2026-05-30, 13:56:

Yeah, but the pci clock comes from the same clockgen chip that does the cpu and ram. I seem to recall SBs needing a secondary cr […]
Show full quote

Yeah, but the pci clock comes from the same clockgen chip that does the cpu and ram. I seem to recall SBs needing a secondary crystal even if the result lines up with pci/x.
In any case, the board I saw a pic of without isa had one next to the chipset so I assume ops does too.

Op take a look in my sig for the bx reference design.
Might give some ideas.

What do you mean with a second crystal, where have you heard about Sound Blasters requiring that?

Anyways, my board doesn't have empty footprint for a crystal. And looking at picture of Abit KT7A, I don't see extra crystals there either, and that board has an ISA slot out of the box.

https://theretroweb.com/motherboards/s/abit-ab-kt7a

.

myne wrote on 2026-05-30, 17:55:

.

That's just real time clock's 32.768kHz oscillator. It doesn't have anything to do with ISA as far as I know, and that's present on my board anyways.

sizzlinbeef wrote on 2026-05-27, 20:17:

Chkcpu wrote on 2026-05-27, 17:44:

Did you already find-out what the undocumented 2.2V setting is on the MS-5136? If not, and if you have a multimeter, I can help […]
Show full quote

Did you already find-out what the undocumented 2.2V setting is on the MS-5136?
If not, and if you have a multimeter, I can help you with a procedure to test what the undocumented JV6/JV7/JV8 jumper settings do.

However, this MS-5136 board uses a linear voltage regulator for the Vcore supply, which already gets very hot when using a Pentium MMX 233. A K6-2 draws a lot more current and can quickly overheat and burn the regulator.
A K6-2+ is a lot more energy efficient and may work here. But for that we need to find an undocumented 2.0V or 2.1V setting. 😉

I did not find out the settings. I do have a multimeter and would be willing to figure them out with your help so we can have them for posterity. I don't have a k6-2+ or 3+ unfortunately so it'll just be an information gathering exercise if you think the board would be at risk with my regular k6-2.

Hi sizzlinbeef,

Here are some hints for the MS-5136 Vcore selection experiment. For this you need to use a dual-voltage CPU that is supported by the board, like a 2.8V Pentium MMX, a 2.9V/3.2V K6 model 6, or a 2.9V Cx686MX.
If during the experiment the Vcore voltage goes up instead of down, these CPUs won’t be harmed during the short test period.

Start by setting the boards jumpers for the installed CPU, however use a low multiplier/FSB setting like 2x50 so that the CPU won’t stall during a low Vcore setting and influence the measurement.
Set the multimeter to voltage DC and connect the black lead to a ground point like one of the 4 middle pins of the AT power connector or the metal housing of the keyboard or mouse port connector.

Switch the PC on and check if it runs correctly. Then use the red multimeter lead to probe the metal lip of the power transistor with the large heatsink. There are two, one for Vio and the other for Vcore.
On one of them you should measure the 2.8V or 2.9V you set with the JV jumpers and if so, you have found the Vcore regulator and the measure point for the experiment.

Now, from the motherboard manual I see the CPU Voltage selection is controlled by eight jumpers JV1-JV8.
JV1 and JV2 clearly control the selection for single- or dual-voltage CPUs. These two jumpers should aways be open for dual-voltage CPUs.
The function of jumper JV3 is unclear, but I should leave if at the 2-3 position as indicated for dual-voltage CPUs.
Jumpers JV4 and JV5 select the 3.38V or 3.52V for single-voltage CPUs and should both be left open for dual-voltage CPUs.
That leaves the Vcore jumpers JV6/JV7/JV8 for the experiment.

First close only JV6 and check if you indeed get 2.5V Vcore.
Then also close JV7 and check the voltage. Depending on how the JV6/JV7/JV8 resistor network is connected, you will see the Vcore either go down or go up.
Repeat the experiment with all other JV6/JV7/JV8 combinations.

If the voltage goes up with more than one JV6/7/8 jumper closed, a final test with all three JV6/7/8 jumpers open may reveal the lower Vcore you want to find.

I hope this procedure is clear and I’m curious what you will find.

Happy testing,
Jan

Mike_ wrote on 2026-05-28, 10:54:

Hi Jan,

I have tried adding an ISA slot to a GA-7ZXE rev 2.1 board which has footprint for one, but I have run into problems and I'm wondering if BIOS lacks some functionality that is needed for ISA. So far I have enabled some DMA configs for the BIOS with AMIBCP and added a missing passive component to board (DACK5 was not connected to pin on ISA slot at all, adding a resistor or piece of wire was needed). Now AWE64 gets correctly assigned to IRQ 5 and DMAs 1 and 5. However it's still not functional, Creative's diagnose.exe correctly recognizes the IRQ and DMA settings and is able to play some garbled audio, but CTCM complains about incorrect Vendor ID and Windows gives Code 9 error.

I have been using version FB BIOS downloaded from Gigabyte's site, it and the version modded with AMIBCP are in the attached zip file.

Hi Mike_,

I see you already got a lot of help with the ISA slot issue on your GA-7ZXE Rev 2.1.
You guys covered a lot of ground concerning a possible hardware issue.

Looking at the BIOSes you uploaded, I see that you correctly enabled the hidden PnP options for all DMA channels. Curious that DMA channel 5 was fixed at ISA/EISA and therefore removed from the PnP pool. On the other DMA channels, PnP was enabled.
This must have something to do with DACK5 being not connected to the ISA slot.

Thinking about this, it looks like Gigabyte needed DMA 5 exclusively for an on-board function and removed it from the PnP pool and ISA bus signals for a slot that wasn’t installed on this Rev 2.1 board anyway.

Searching the VT82C686B datasheet for a function that could use DMA 5, I noticed the build-in digital audio controller in this southbridge (function 5). This controller should be AC97 and SB Pro compatible and provides on-board audio with the help of an ALC201A chip that houses the DAC and mixers.

In the BIOS I see several options to control this on-board audio and only the ‘Onboard Sound’ option is hidden and always enabled. Maybe this function is interfering with your AWE64?
Making this option visible through AMIBCP may help with fixing your issue.

Happy tinkering,
Jan

Chkcpu wrote on 2026-05-30, 20:35:

Hi Mike_, […]
Show full quote

Hi Mike_,

I see you already got a lot of help with the ISA slot issue on your GA-7ZXE Rev 2.1.
You guys covered a lot of ground concerning a possible hardware issue.

Looking at the BIOSes you uploaded, I see that you correctly enabled the hidden PnP options for all DMA channels. Curious that DMA channel 5 was fixed at ISA/EISA and therefore removed from the PnP pool. On the other DMA channels, PnP was enabled.
This must have something to do with DACK5 being not connected to the ISA slot.

Thinking about this, it looks like Gigabyte needed DMA 5 exclusively for an on-board function and removed it from the PnP pool and ISA bus signals for a slot that wasn’t installed on this Rev 2.1 board anyway.

Searching the VT82C686B datasheet for a function that could use DMA 5, I noticed the build-in digital audio controller in this southbridge (function 5). This controller should be AC97 and SB Pro compatible and provides on-board audio with the help of an ALC201A chip that houses the DAC and mixers.

In the BIOS I see several options to control this on-board audio and only the ‘Onboard Sound’ option is hidden and always enabled. Maybe this function is interfering with your AWE64?
Making this option visible through AMIBCP may help with fixing your issue.

Happy tinkering,
Jan

Actually I also noticed that hidden setting after I had posted the ROM files, but unfortunately it didn't help.

But I got an idea what might be causing this on hardware side. Near south bridge there are two IC footprints, but there are only two resistor nets there instead.

Those resistor nets are between ISA slot and south bridge on address lines 0-15, but 470Ω is awfully lot of series resistance even for a slow bus like ISA. On 686B south bridge those lines are multiplexed with secondary IDE's data lines, so maybe the point is to prevent ISA bus messing with whatever is happening on secondary IDE. Looking at pictures of GA-7ZX, which has ISA slot as an option, there are two TI's "F245" chips.

https://theretroweb.com/motherboards/s/gigabyte-ga-7zx-1.x

I suppose it could be this? Locations of GND and Vcc pins match, and OE on both sockets is low when board is running, ie. it's enabled. I guess Gigabyte figured out that a cheap resistor net would suffice for reading from a slow BIOS chip during startup, and there's no need to have an actual IC there. 😁

https://www.ti.com/lit/ds/symlink/sn54f245.pdf

That looks like a winner.
Now the fun part. Finding some.
Or bridging them and disabling ide2?

myne wrote on 2026-05-31, 08:39:

That looks like a winner.
Now the fun part. Finding some.
Or bridging them and disabling ide2?

Not much "fun" in finding those, as you can just buy them from Digikey or TME. 😀

Chkcpu wrote on 2026-05-30, 19:51:

First close only JV6 and check if you indeed get 2.5V Vcore. Then also close JV7 and check the voltage. Depending on how the JV6 […]
Show full quote

First close only JV6 and check if you indeed get 2.5V Vcore.
Then also close JV7 and check the voltage. Depending on how the JV6/JV7/JV8 resistor network is connected, you will see the Vcore either go down or go up.
Repeat the experiment with all other JV6/JV7/JV8 combinations.

If the voltage goes up with more than one JV6/7/8 jumper closed, a final test with all three JV6/7/8 jumpers open may reveal the lower Vcore you want to find.

I hope this procedure is clear and I’m curious what you will find.

Happy testing,
Jan

Yes, jumpering JV6 gets the correct 2.5v.

6-7 closed 1.93v
6-7-8 closed 1.73v
7-8 closed 2.04v
6-8 closed 1.95v

Unfortunately I can't confirm any of these result in a working system. I tried booting this board before testing. Using default settings for a pentium MMX (I tried a 166, 200, 233), can't get the thing to post. No video output. Tried different VGA, RAM, deoxit on all slots, alcohol clean, etc. I think time has taken it's toll on something on the board. It may be related to getting 3.6v on the VIO transistor when it should be 3.3v with JV1 and 2 open and JV3 set to 2-3. It has been probably a decade since I last tried booting it, so not surprising.