Stainlesscat wrote on 2021-03-24, 16:31:

My presented topic was about the presence of compatible LPC to PCI bridges to get Sound Blaster ISA resources (ISA I/O port resources not DMA!) in order to get working emulation of sound blaster ports to work because ICH7 and later only support ISA style port resources with legacy ports decoding through PCI with conjunction by a compatible LPC bridge connected to ICH chipset, The chipsets cannot handle ISA resources alone which is why all of them have an LPC Bus present with an LPC bridge connected. I am not referring to chipset exclusive features support like D-DMA or PC-PCI! as others stated already that ICH5 was bottom of the road... D-DMA is an ISA DMA resource exclusive to certain chipsets that supported it and i'm well aware Intel's I/O Controller Hub never supported this.

And Just if there is confusion about what i'm talking about let me be clear... I'm only referencing chipsets like ICH7 through 10 as those chipsets do have Legacy ISA through PCI resources but not ISA DMA resources (ISA I/O ports only). And again i'm not refering to exclusive support like PC-PCI (SB-LINK). Also note LPC to ISA bridges are only important actual ISA devices which my topic was not about. Mines was about LPC to PCI chipsets which does give ISA resources (Again no DMA) through PCI which is needed for SOUND BLASTER PORT RANGE FOR SOUND BLASTER EMULATION!

The whole reason i wrote my previous posts because i found out that sound cards that use Sound Blaster TSR driver emulation can still not work without the previously aforementioned requirements.

Okay, sorry about the misunderstanding. It seems some boards in that range (like yours) have trouble getting even basic stuffs (excluding DMA) to work.

IMO the Winbond chip (maybe ITE, sometimes Fintek) that I can always find on a motherboard is a Super I/O. Just took a look at W83627HF (which is one of the common ones) and it does also appear to provide a complete LPC interface along with other stuffs (PS/2, COM, LPT, etc.), so you're right about the LPC bridge thing.

It's just that I don't have a clear connection between PCI sound card's legacy emulation and the chipset's LPC interface yet. I tried looking for one, but nothing looked clear to me as some motherboards exhibit similar behaviors for a given sound card, while they gave me completely contradictory information (maybe DMI is not the right place to look for certain details).

Don't know which chip the board you're having problems with uses, though. If the board is using the same chip as others (not crippled ones) then it's possible the BIOS might have disabled some functionalities that the board is not actually using (or maybe not wired at all). After all, OEM boards tend to lock down a lot of stuffs, which I think is to ensure users don't accidentally void their warranty.

On the other hand, IRQ handling can be problematic with modern motherboards. I recall having a very hard fight getting it right on new VIA chipsets. PCI audio cards in question would not work unless I make the card actually take IRQ 5 through trial and error (yeah, sometimes I have to install expansion cards to particular slots and even in a particular order to get it right). Reserving IRQ5 for legacy usually don't work in such cases, but it helped getting ISA sound cards working on older VIA chipsets...

Removed my previous post, as the questions I guess I could have answered on my own by rereading the thread with attention. @stainlesscat does your research translate to (in layman's terms) that for MBs with modern chipsets legacy IO ports are in fact emulated by dedicated controller that allows hw like PS/2 keyboards and mice work, but would only support a subset of IO ports that's needed for that, and that using arbitrary IO port like the well known SB ports may not be supported at all?

Hello guys, I'm kindly asking for your help as we've got totally stuck with our Fortex project.
So, Fortex is basically a combo of YMF744 and AU8830 chips behind a PCI-PCI bridge. For our project we've chosen TI PCI2050BPDV and Intel 21150BC bridges. The project idea was to use the Yamaha part for everything legacy-related (SB emulation, OPL3 FM, MPU-401) in DOS and for MIDI in Windows, while the Vortex part should be used for A3D 2.0 games. Long story short, in Windows everything works as intended, but in DOS the legacy mode does not work. I have yet to try the Vortex part, but with YMF744 I get the following results:
1. SETUPDS gives me errors: "Can not find Sound Blaster at 0x220" when I try to enter Sound test, Native audio 16bit sound plays just fine, FM Sound test gives me error "Invalid File (C:\DS-XG\YMH.MID)" (while the file is obviously fine). Everything plays just fine if I simply put an AOpen YMF744 card without any bridge in the middle.
2. When I try to change the YMF744 registers so that it gets into the "native" I/O range, the FM chip becomes accessible in AT2, plays just fine (while OPL3 is at 0xA860h). However I still couldn't make Duke3D work with editing the DUKE3D.CFG manually.

I tried to test the same "combo" with production cards (AOpen YMF744 and Aureal SQ2500) into a production Intel PCI riser card (based on TI PCI2031 bridge chip). Moreover, I tried modifying the PCI2031 registers to enable the subtractive decode. When I do it on the secondary bus, nothing changes. When I do it on the primary bus, the system makes a short beep and keyboard stops responding (however the cursor is blinking so the system is not on a total halt).

Any help will be very much appreciated.

Cursor blinking is a video hardware feature and it'll blink regardless of what CPU is doing. If numlock etc. LEDs still respond the system isn't totally crashed but if they don't, something has gone very wrong...

But other than that, I have no ideas about what can help here. I assume YMF744 on its own, without any bridges, works properly on the test machine ?

LEDs don't respond.

YMF744 without any bridges works totally fine.

Is there another bridge on that system? You cannot enable subtractive decoding on more than one.

It's an ASUS XG-DLS system with 440GX chipset, no other bridges apart from the north bridge and south bridge (PIIX4E) + the PCI2031 in question.

I guess something goes wrong with legacy IO ports forwarding on the path via bridges to soundcard. Probably need to configure some IO window with proper address. Enabling subtractive decode may cause that it sinks all legacy IO and so system stops responding. I don't know about priorities what happen if more bridges has sub.decode enable. Probably one is PCI2LPC bridge that connects superIO...

I've come to the same thoughts about IO windows as PCI2031 mentions this capability in the datasheet...

If you cannot use subtractive decoding, you can try to program i/o ranges 0x220 and 0x388 on the two extension windows that the bridge provides

fjardest wrote on 2023-11-01, 22:45:

If you cannot use subtractive decoding, you can try to program i/o ranges 0x220 and 0x388 on the two extension windows that the bridge provides

Yes, I could try this if possible.

Hello guys, I'm totally stuck. Just wrote a post in the Fortex thread (Re: Fortex, the A3D & XG/OPL3 accelerator (Vortex 2 + YMF744 combo sound card)), if any of you can help me out, that would be great. Fingers crossed.