retardware wrote on 2021-10-17, 19:07:

I do not find it anymore in my history, and Kontron seems to have redone their site.
It was mentioned in an article about a particular board, with BIOS screens etc how ISA stuff gets set up, including IRQs and DMA.

The only thing I still found is this whitepaper: https://www.kontron.com/resources/collateral/ … paper_pc104.pdf

They talk about ISA starting page 3. How they specifically chose the amd A55E chipset/hub for the integrated PCI controller allowing access to the lower memory map and configured to " have all unused PCI
cycles are mapped to ISA" and on page 6 there is mention of the implementation of Distributed DMA to get the ISA DMA working, allowing configuration in the BIOS.

Seems to be a system from around 2011.

Very interesting read.

weedeewee wrote on 2021-10-17, 16:08:

The chip in question, fintek f85226 is LPC to ISA Bridge. LPC seems to support DMA transfers […]
Show full quote

LSS10999 wrote on 2021-10-17, 15:30:

weedeewee wrote on 2021-10-17, 08:16:

Isn't it possible to have some software handle the DMA transfers. The chipsets in their pdfs state they actually support dma transfers and it's not like the pcs are too slow to handle it since the dma used to run on an... 8MHz bus.

The bus in question (LPC) has been superseded by a new one: eSPI.

This one doesn't support ISA-style DMA that standard floppy disk controllers expect.

The chip in question, fintek f85226 is LPC to ISA Bridge.
LPC seems to support DMA transfers

maybe I should rephrase my question...

Can software be written that listens to/controls these LPC to ISA or PCI to ISA bridges for the handling of, the outdated and no longer directly in hardware supported due to the too many different layers of busses the data needs to go through, DMA transfers ?

The LPC-ISA bridge has functional ISA DMA, as demonstrated on RUBY-9719VG2AR, if configured.

it's just that the PCHs since 100 series are switching to eSPI. While the chipset can still expose a LPC bus, it won't support DMA. If you read the chipset's datasheet you'll notice the absence of pins like LDRQ, LDCK.

retardware wrote on 2021-10-17, 19:07:

I do not find it anymore in my history, and Kontron seems to have redone their site.
It was mentioned in an article about a particular board, with BIOS screens etc how ISA stuff gets set up, including IRQs and DMA.

The only thing I still found is this whitepaper: https://www.kontron.com/resources/collateral/ … paper_pc104.pdf

I've read that before. And according to this (which I think it means the PDF you referred to), it seems to be a firmware-based virtual implementation probably leveraged SMM similar to VSA. As I haven't seen actual devices I can't know for sure...

LSS10999 wrote on 2021-10-18, 00:52:

...it seems to be a firmware-based virtual implementation probably leveraged SMM similar to VSA. As I haven't seen actual devices I can't know for sure...

Yes exactly. The article I still haven't found again talked about this and that this was the part that took most development resources.
Taking the "request" from the ITE chip, and then modifying it to pass the DMA functionality through PCI and then PCI-E via some BIOS "middleware".

While searching for that article again, I came across an interesting discussion on the Intel board, where also some funny details are told... it looks like that Intel engineers got stuck some time until they found that it was a bad ISA connector (probably too modern production) which made their tests inexplicably fail. So there are a lot of potentially unexpected pitfalls trying to get ISA work 😀

retardware wrote on 2021-10-18, 01:41:

Yes exactly. The article I still haven't found again talked about this and that this was the part that took most development res […]
Show full quote

LSS10999 wrote on 2021-10-18, 00:52:

...it seems to be a firmware-based virtual implementation probably leveraged SMM similar to VSA. As I haven't seen actual devices I can't know for sure...

Yes exactly. The article I still haven't found again talked about this and that this was the part that took most development resources.
Taking the "request" from the ITE chip, and then modifying it to pass the DMA functionality through PCI and then PCI-E via some BIOS "middleware".

While searching for that article again, I came across an interesting discussion on the Intel board, where also some funny details are told... it looks like that Intel engineers got stuck some time until they found that it was a bad ISA connector (probably too modern production) which made their tests inexplicably fail. So there are a lot of potentially unexpected pitfalls trying to get ISA work 😀

This is an archived link for the article mentioned in FlaterCo. However, the whitepaper itself is not archived (I think it's not publicly available).

As for the discussion, it's not really about ISA DMA, but ISA in general (the card in question uses only I/O address).

The discussion is very interesting, but I am not sure that I understand everything unfortunately!

Just to summarize, if I understand correctly, there is a chance that a floppy controller with bios might work with the CS620-H310 or the MS-98L9 V2.0. One way would be that it is standard for a floppy controller to revert to non-dma mode of dma is not working/present? The other way would be that these boards do support DMA despite the evidence to the contrary, but there is no way to know for sure unless I try?

The primary reason for me to get one of these boards is to have real internal non-usb floppy drives 5.25 and 3.5 on the most modern board as possible. If the floppy controller doesn't work, what are the advantages of these boards vs. a board with PCI but not ISA? Some sound cards that require dma would not work either. What legacy hardware would be interesting to install in such a board?

Thanks!

wondow wrote on 2021-10-21, 01:21:

The primary reason for me to get one of these boards is to have real internal non-usb floppy drives 5.25 and 3.5 on the most modern board as possible. If the floppy controller doesn't work, what are the advantages of these boards vs. a board with PCI but not ISA? Some sound cards that require dma would not work either. What legacy hardware would be interesting to install in such a board?

Thanks!

For 3.5", technically speaking, an LS120 and a PCIe IDE controller (Marvell preferably) would be the way to go to get one on the "newest" system...

There do exist "PCI" floppy controllers... I have one with a UM8663 chip. But don't get your hopes up - I tested it in a P4 board and it doesn't work. I think this card was meant to supplement a very specific motherboard, because when plugged into this more modern board, the PS/2 ports stop working. Either that or the QFP chip needs to be replaced on the card, but that's doubtful.

If the UM8663 could somehow be made to work with modern boards then something with a PCI to PCIe bridge should accommodate it.

Its quite long time, what about than Dos sound test?

Has anyone checked whether the PCI sound card Yamaha 754 with VXD driver under Windows 98 will work on MS-98L9 V2.0? I see the same PCI-e to PCI ITE bridge on this board that I recently had problems with on another board and am considering replacing the board with something else like this one. But you need to know for sure whether it makes sense (Hope for a more even BIOS from MSI) or will everything be the same? Who has MS-98L9 V2.0, would you mind testing Yamaha 754 with VXD driver under Windows 98?

DoZator wrote on 2022-08-14, 19:30:

Has anyone checked whether the PCI sound card Yamaha 754 with VXD driver under Windows 98 will work on MS-98L9 V2.0? I see the same PCI-e to PCI ITE bridge on this board that I recently had problems with on another board and am considering replacing the board with something else like this one. But you need to know for sure whether it makes sense (Hope for a more even BIOS from MSI) or will everything be the same? Who has MS-98L9 V2.0, would you mind testing Yamaha 754 with VXD driver under Windows 98?

I don't own this board but there was a similar discussion before.

I think Yamaha's VxD driver probably requires its legacy block be properly initialized in order to work, whereas WDM driver does not (and thus it does not provide any legacy functionalities).

I'm not sure why you need VxD driver. If you need FM for MIDI, you can try using Microsoft's WDM driver which was included in Windows ME.

PS: Actually Yamaha's NT4 driver does the same. The driver was supposed to be usable on NT 3.51 also, as tested on a 440BX-based laptop that had it integrated, but on my target AMD motherboard (M5A78L/USB3), the driver couldn't be loaded, and returned me an error code. Further inspection through disassembly revealed that the particular error code was due to it unable to initialize the legacy block (which was expected as the chipset in question doesn't support any form of legacy audio). I tried bypassing the legacy audio block initialization by modifying some instructions in the driver so it could be installed and loaded, but the moment it tried to play any audio the NT system hardlocked.

Even if it did work, this board is so new that it would have severe ACPI, chipset, and other motherboard resource problems. But for reference, I already had a lot of problems trying to get a Yamaha PCI card working using VxD drivers working on an Intel 945 chipset. Aureal Vortex and any of the Creative PCI Sound Blasters and Audigy cards I tried worked fine though, and ask of these provide some kind of legacy compatibility.

Since it has an ISA slot, use that instead for Adlib OPL instead of trying to get it from a Yamaha PCI card. That's what the port is most useful for on these motherboards anyways.

Kahenraz wrote on 2022-08-15, 08:26:

Even if it did work, this board is so new that it would have severe ACPI, chipset, and other motherboard resource problems. But for reference, I already had a lot of problems trying to get a Yamaha PCI card working using VxD drivers working on an Intel 945 chipset.

Also for reference, on my everyday PC Ryzen 3900X/Biostar X470 motherboard I have PCI audio on all Windows versions using 2 different PCI sound cards:

ALS4000 - doesn't work in Windows 3.1 but works in every other OS - including NT 3.51/4
Ensoniq ES1370 (not the Creative version) - doesn't work in NT 3.51 but works in every other OS, including Windows 3.1 (actually works great in 3.1, including MIDI)

I stick with Ensoniq card because I'm a big fan of Windows 3.11 - and I'm still amazed that I have drivers for everything - sound, video (using a Voodoo 3 PCI) and LAN.

Kahenraz wrote on 2022-08-15, 08:26:

But for reference, I already had a lot of problems trying to get a Yamaha PCI card working using VxD drivers working on an Intel 945 chipset.

Did you end up succeeding? Perhaps the description of the solution to the problem for the Intel 945 will also help to cope with the problem on later chipsets, or at least push in the right direction. However, it is not yet clear whether the Yamaha 754 will work on the MS-98L9 V2.0 with the VXD driver under Windows 98 or not. This can only be said for sure by someone who has already tried it.

I don't know. Some methods used to acquire legacy DMA seem to not work well with certain PCI should cards on newer chipsets. What I'm saying is that I've found the Yamaha PCI cards to be particularly susceptible to this problem. My ESS Solo-1 also has problems with my i945. In Windows, it doesn't work at all and causes the system to lock up.

Unfortunately, it's the Yamaha and ESS PCI cards that have the best Adlib OPL representation, and these tend not to work. This is why having at least one ISA slot for Yamaha OPL is such a luxury. It doesn't require DMA and always works, even on newer chipsets.

Hello!

Has anyone been able to run and test DMA on the MS-98L9 V2.0 board ?

I want to make the disk drive controller work on this board.
https://texelec.com/product/quad-flop-four-po … ppy-controller/

Remember EISA? These motherboards actually come with UISA slots, where "U" stands for useless. 😀

Well, almost useless. The latest model from DFI comes with an 82-page manual that walks you through setting up a combo of Ubuntu + QEMU + XP in a VM, and now you should be good to use your ISA card. I bet this isn’t exactly what anyone imagined when they saw those ISA slots...

"DFI provides a virtualization solution that allows new x86 platforms to use ISA devices. The CS620 includes a host image (Ubuntu) with the KVM hypervisor, letting users install their legacy OS. This document guides users through legacy OS installation and ISA device configuration."

I happen to have the previous DFI model ( HD620-H81 ). This one doesn’t include a custom ISO, and there’s no manual explaining any VM setup. In theory, it should work on bare metal, but no one has ever managed to get those ISA slots to do anything useful. I couldn’t even get a simple scanner interface card to be recognized. To back this up, I’ve attached a comment from a German forum discussing the issue.

OMORES wrote on 2025-04-01, 19:14:

Remember EISA? These motherboards actually come with UISA slots, where "U" stands for useless. :) […]
Show full quote

Remember EISA? These motherboards actually come with UISA slots, where "U" stands for useless. 😀

Well, almost useless. The latest model from DFI comes with an 82-page manual that walks you through setting up a combo of Ubuntu + QEMU + XP in a VM, and now you should be good to use your ISA card. I bet this isn’t exactly what anyone imagined when they saw those ISA slots...

"DFI provides a virtualization solution that allows new x86 platforms to use ISA devices. The CS620 includes a host image (Ubuntu) with the KVM hypervisor, letting users install their legacy OS. This document guides users through legacy OS installation and ISA device configuration."

I happen to have the previous DFI model ( HD620-H81 ). This one doesn’t include a custom ISO, and there’s no manual explaining any VM setup. In theory, it should work on bare metal, but no one has ever managed to get those ISA slots to do anything useful. I couldn’t even get a simple scanner interface card to be recognized. To back this up, I’ve attached a comment from a German forum discussing the issue.

Those DFI boards use PCIe-PCI and PCI-ISA bridge combo to provide ISA slots.

For post-ICH5 chipsets, DMA cannot be used this way, but should allow using any available I/O port and IRQ, unless PCIe imposes more restrictions than PCI.

As for the attached post... all those use cases (IDE, FDC) involve DMA so there's no way to get it work. Use cases that don't use DMA should work okay, including sound card's FM synth, or directly accessing the sound card's I/O port for SFX.

Considering the manual regarding how to use ISA devices inside WinXP VM guests, I think the board is probably designed in a way that the PCIe-PCI bridge (IT8892) before PCI-ISA bridge is in its own dedicated IOMMU node and may be trivially configured for passthrough so the guest can access it as well as the PCI-ISA bridge and ISA devices connected to it.

I understand how DMA works, and I know it's not available on these boards, though it's ironic that you can manually wire into the LPC bus on some pretty modern motherboards to get DMA, while an expensive ISA board like mine can't do anything about it, even though it was technically possible. But, yeah this workaround might not be suited for industrial use.

My point is that these modern ISA motherboards should be avoided, even for non-DMA use cases. I couldn't get any FM Synth functionality to work on any ISA sound cards tested. The BIOS is extremely barebones, with no effort put into enabling useful features. There are no ISA/PCI-related settings in the BIOS, though if you examine the BIOS with an editor, you'll find hidden options like 'subtractive decoding' and a few others that might help, but they’re all disabled. Even IDE legacy mode is disabled.

Is it possible to fully use the ISA slot on a virtual machine?

I tried to put this controller on the MS-98L9 V2.0 board.:
https://texelec.com/product/quad-flop-four-po … ppy-controller/
Even the built-in BIOS does not start, and I have no idea how to start this board.