tecpossum wrote on 2026-07-06, 22:04:

In case someone finds it useful, a full face on shot of an assembled 0.4 board, as I don't think one has been posted: […]
Show full quote

In case someone finds it useful, a full face on shot of an assembled 0.4 board, as I don't think one has been posted:

The attachment disappointment-v0.4-assembled.jpg is no longer available

And the obligatory "it works" shot with a PicoGUS and Doom in the background:

The attachment hey_it_works.jpg is no longer available

Initial testing was done with a H61MXV as someone was selling them on fleabay for 20 squids a pop with CPU and RAM, and it seemed a safe bet. Hopefully I'll find time this weekend to test some other candidates I have lying around.

Huge thanks to rasteri for making this project and releasing it.

Awesome!

Also I'm impressed with the number of PCBs in this photo

Now that you mention it, that is quite the stack.

Probably silly question, but can it support more than one card? Would it be illogical to make it into an isa card that slots into a backplane?

myne wrote on 2026-07-08, 04:36:

Probably silly question, but can it support more than one card? Would it be illogical to make it into an isa card that slots into a backplane?

The F85226 should be able to handle up to 3 ISA slots. Tiido has a design some pages back.

For dISAppointment, I think ISA risers/extenders can be used to provide additional slots.

tecpossum wrote on 2026-07-06, 22:04:

And the obligatory "it works" shot with a PicoGUS and Doom in the background:

PicoGUS in GUS mode also works ? In 0.2 adapter GUS mode works only in port mode, no DMA (Impulse tracker). Duke, Doom, Descent - no sound and/or hang.

vsharun wrote on 2026-07-08, 18:16:

PicoGUS in GUS mode also works ? In 0.2 adapter GUS mode works only in port mode, no DMA (Impulse tracker). Duke, Doom, Descent - no sound and/or hang.

That was in SB16 mode, and it had to be on IRQ7. IRQ5 resulted in no sound, even although every tool I tried said it was free. My assumption is the SuperIO swallowed it even with LPT disabled.

Didn't test GUS, but I can add it to the list for the weekend.

tecpossum wrote on 2026-07-08, 21:42:

vsharun wrote on 2026-07-08, 18:16:

PicoGUS in GUS mode also works ? In 0.2 adapter GUS mode works only in port mode, no DMA (Impulse tracker). Duke, Doom, Descent - no sound and/or hang.

That was in SB16 mode, and it had to be on IRQ7. IRQ5 resulted in no sound, even although every tool I tried said it was free. My assumption is the SuperIO swallowed it even with LPT disabled.

Didn't test GUS, but I can add it to the list for the weekend.

Fascinating, thanks. I'll check new 4.1.1 firmware this weekend.
Also hit the wall with IRQ5 few weeks ago.
Have you managed to run SB16 mode in Duke3D ? Or in SB Pro mode only ? In other words: high DMA, any issues ?

Duke3D + SB16 with setup.cfg hack to put both DMAs on 1 works on second launch, first launch SFX is severely distorted.

Can't get any sound in GUS mode with the standard forwarded ranges. I'll need to pull a proper ISA motherboard to make sure I'm not doing something stupid and/or try get the rp2350 ISA analyzer firmware on my DIY PicoGUS running to see if there's anything interesting happening.

tecpossum wrote on 2026-07-08, 21:42:

That was in SB16 mode, and it had to be on IRQ7. IRQ5 resulted in no sound, even although every tool I tried said it was free. My assumption is the SuperIO swallowed it even with LPT disabled.

Didn't test GUS, but I can add it to the list for the weekend.

Did you check the PIRQ registers of the LPC controller (60h and 68h of 00:1f.0)?

If any byte reads 05h or 07h it means IRQ5 or 7 is being routed to onboard or discrete PCI/PCIe devices.

In that case you'll need to check which PCI device is taking that IRQ. If it's a non-essential onboard device, disable it from BIOS.

tecpossum wrote on 2026-07-11, 01:02:

Duke3D + SB16 with setup.cfg hack to put both DMAs on 1 works on second launch, first launch SFX is severely distorted.

Can't get any sound in GUS mode with the standard forwarded ranges. I'll need to pull a proper ISA motherboard to make sure I'm not doing something stupid and/or try get the rp2350 ISA analyzer firmware on my DIY PicoGUS running to see if there's anything interesting happening.

You may try Impulse tracker - works. Which proves DMA and ports are okay. Dunno what to do next.
I suppose demoscene will be okay as well.

tecpossum wrote on 2026-07-11, 01:02:

Can't get any sound in GUS mode with the standard forwarded ranges. I'll need to pull a proper ISA motherboard to make sure I'm not doing something stupid and/or try get the rp2350 ISA analyzer firmware on my DIY PicoGUS running to see if there's anything interesting happening.

Should note that GUS mode expects more I/O ports in the 300h range.

Make sure ports 3X0h-3X7h (X is the same as your GUS base port 2X0h, so if your base port is 240h then it's 340h-347h). So make sure the mask for the range meant for 300h-370h range is at least 74h.

Originally the mask was set to 70h by default which means 3X0h-3X3h, enough for MIDI but not enough for GUS. I've since updated lpcisa (in my lpcexp) to default to the new range.

I was able to (finally!) find some time to look at my X10SAE board in more detail.

Unfortunately the supposed LDRQ pin on the TPM header is indeed not connected.

Without a boardview it’s hard to source LDRQ directly. And lifting a SuperIO pin and soldering to the underlying pad is definitely beyond my soldering skill level 😂

So I think this board is a dead end.

Were we ever able to find a board with fully routed LDRQ? Or perhaps something that’s relatively easy to mod? 🙏

jamesfmackenzie wrote on Yesterday, 09:19:

Were we ever able to find a board with fully routed LDRQ? Or perhaps something that’s relatively easy to mod? 🙏

AFAIK no board had ever actually routed LDRQ1# as it's not of any real use for TPMs. However, on most boards LDRQ1# is routed to an unpopulated resistor footprint so you can solder a wire to make it accessible, preferably through a NC (unconnected) pin of an existing header.

You'll definitely need boardviews. Schematics can be of some help but not as much as boardviews, as you still need to manually look at the whole board to locate the footprint in question.

LSS10999 wrote on Yesterday, 10:52:

jamesfmackenzie wrote on Yesterday, 09:19:

Were we ever able to find a board with fully routed LDRQ? Or perhaps something that’s relatively easy to mod? 🙏

AFAIK no board had ever actually routed LDRQ1# as it's not of any real use for TPMs. However, on most boards LDRQ1# is routed to an unpopulated resistor footprint so you can solder a wire to make it accessible, preferably through a NC (unconnected) pin of an existing header.

You'll definitely need boardviews. Schematics can be of some help but not as much as boardviews, as you still need to manually look at the whole board to locate the footprint in question.

Thanks for the info!

Agree that the boardview is incredibly helpful. For my Z87-PRO it instantly revealed a single pull-up resistor for LDRQ1# (although it's under the PCH heatsink -- but that's another story 😂). Would be great to have the same for the X10SAE -- since it already has PCI slots and would make a great Time Machine. I'm sure there's a resistor or resistor footprint somewhere I could use 😎

jamesfmackenzie wrote on Yesterday, 09:19:

Unfortunately the supposed LDRQ pin on the TPM header is indeed not connected.

Is it not connected to anything at all?

Probably time to update this.

Anyone want to make it into a gsheets?

On another note, did anyone ever figure out the AMD side of things?
I recently realised AM4 is a SOC. It doesn't technically even need a chipset.
In theory, you could design a "retro board "smaller than ITX or at least PC-104 that fit an APU, Pi on LPC, sata/nvme M.2, an LPC or SPI attached bios and practically nothing else.
Most would want a PCIE slot for the option to run a GPU but that's only one slot.
I haven't looked into FM2, but if that's the same, there's a retro-project I think we could realistically achieve here.
The down side is that AMD datasheets are impossible to find.
The chips are dirt cheap though.

jamesfmackenzie wrote on Yesterday, 11:20:

LSS10999 wrote on Yesterday, 10:52:

jamesfmackenzie wrote on Yesterday, 09:19:

Were we ever able to find a board with fully routed LDRQ? Or perhaps something that’s relatively easy to mod? 🙏

AFAIK no board had ever actually routed LDRQ1# as it's not of any real use for TPMs. However, on most boards LDRQ1# is routed to an unpopulated resistor footprint so you can solder a wire to make it accessible, preferably through a NC (unconnected) pin of an existing header.

You'll definitely need boardviews. Schematics can be of some help but not as much as boardviews, as you still need to manually look at the whole board to locate the footprint in question.

Thanks for the info!

Agree that the boardview is incredibly helpful. For my Z87-PRO it instantly revealed a single pull-up resistor for LDRQ1# (although it's under the PCH heatsink -- but that's another story 😂). Would be great to have the same for the X10SAE -- since it already has PCI slots and would make a great Time Machine. I'm sure there's a resistor or resistor footprint somewhere I could use 😎

Given that supermicro lists pin 20 as ldrq, you might have luck working backwards. Either follow the trace visually and confirm with multimeter, or clip onto 20 and poke empty pads hoping for a beep

The front has some candidates

myne wrote on Yesterday, 23:46:

Probably time to update this. […]
Show full quote

Probably time to update this.

Anyone want to make it into a gsheets?

On another note, did anyone ever figure out the AMD side of things?
I recently realised AM4 is a SOC. It doesn't technically even need a chipset.
In theory, you could design a "retro board "smaller than ITX or at least PC-104 that fit an APU, Pi on LPC, sata/nvme M.2, an LPC or SPI attached bios and practically nothing else.
Most would want a PCIE slot for the option to run a GPU but that's only one slot.
I haven't looked into FM2, but if that's the same, there's a retro-project I think we could realistically achieve here.
The down side is that AMD datasheets are impossible to find.
The chips are dirt cheap though.

AMD does have datasheets for those chipsets. However, the problem is I've no idea whether those datasheets are actually correct, as well as whether there is anything else I need to fiddle with.

Ryzen CPUs have the APU-era FCH integrated into the CPU. Just that they have only one LDRQ# signal (LDRQ0#) exposed and that is usually wired to the SuperIO. It appears some ASUS boards used some kind of stripped-down SuperIO variants that don't use LDRQ# so LDRQ0# appears to be unused there. PRIME B450M-A seems to be like that according to its boardview.

The LDRQ0# signal is no longer present from AM5 and onwards. FM2 and before have both LDRQ0# and LDRQ1# present, though not all boards actually exposed LDRQ1# to an accessible footprint in case of AMD CPU boards, according to a few boardviews I checked...