"...and relief that is almost tidal in scope washes over me in an awesome wave" 🪓😌

ExplodingLemur wrote on 2023-05-01, 15:00:

I just bought an L440GX+ off eBay and I've got a YMF744 sound card to test with. I'll give it a try when the motherboard arrives.

Just a little heads-up, have you received the board already?

hard1k wrote on 2023-06-26, 20:09:

Just a little heads-up, have you received the board already?

Yeah, haven't been able to do anything with it other than mount it to a test stand though 😒 Travel, then COVID, then more travel...will see if I can get it done this weekend.

Great, thanks!

Well, crud 😒 The board is dead. At first it was starting the PSU the moment I applied power but no beep codes or POST activity on a POST card. After swapping some hardware around (RAM, CPU, removing other PCI cards, etc) it won't turn on at all. The power supply is confirmed good with a PSU tester though. Sorry 🙁

Ohhhh that's bad.
We have bad news here too. Both chips don't function properly in any of their legacy modes (despite working perfectly fine natively in Win9x), no SB emulation, no FM, no MPU output, neither in pure DOS, nor in DOS in Windows. I believe it's due to some misconfiguration of the bridge, maybe some registers are set wrong, I don't know. As we don't have any skilled low-level programmer, we're desperately looking for one. We can ship the prototype board anywhere in the world. Help Me, Obi-Wan Kenobi. You're My Only Hope.

My hypothesis is that the legacy DOS functionality of these chips needs a bridge that supports subtractive decode to work properly.

I admit I'm no expert though, so I could be wrong.

Hello guys,

So, we're stuck with the bridge settings apparently.
The legacy part of YMF744 doesn't work in DOS at its intended resources (0x220h / IRQ 5 / DMA 1 for SB, 0x388h for FM). Using the hack from this thread (Using Yamaha the YMF744 under DOS without legacy/ISA addressing) I've managed to get the OPL3 working in AT2 at 0xA860.
While studying the PCI2050 datasheet, I understand that our chosen bridge is incompatible with the original project idea as it does not offer any capabilities for I/O extension windows (while its PCI2250 sister chip does).
However before going into the project revision and sourcing more capable bridge we decided to make a proof of general concept - we've sourced a production batch YMF744 card (AOpen Cobra AW744L II), a production AU8830 card (Aureal SuperQuad SQ2500), and a production Intel 2-slot PCI riser card (based on TI PCI2031 bridge chip that does state the needed capabilities). So, combining these three what we get is essentially a Fortex but 20 times bigger and clumsier but whatever.
Unfortunately my skills and knowledge proved not sufficient to make this combo work in pure DOS despite reading the datasheets and messing with different tools to initialize and patch the registers of these devices.

So, I'm asking you for help - if there is anyone with solid understanding of how these things work at low level, please drop me a line. I have the testing machine here (based on ASUS XG-DLS, thus a bullet proof 440GX chipset), with a Fortex prototype card on hand, and with the mentioned combo of PCI2031+YMF744+AU8830 as production working cards.

Looking forward to hearing from you!

P.S. This project was intended as a non-commercial, just to cover our R&D and production costs when/if released. However if anyone of you want to collaborate on a paid basis, feel free to offer your services too (but bear in mind that in this case the payment will come only as a success fee when we get the beauty running).

Hello,
Have you tried to use extension window 0 for sb, writing 0x220 to base register ( pci config space 44h), 0x23f to limit register ( config 48h) write 0 to bit 0 of config space 65h and finally write 1 to bit 0 of config space 64h. Config space 66h should have the default value of 0x6

Hello,
Maybe using I/O base register ( Offset: 1Ch) and I/O limit register (Offset: 1Dh) to define an I/O address range (100h to xxxh) and ISA Enable bit can be useful.