TwinSlot for 6360 & 5x00 tests being done IRL. 8+ years long project draws to a close!

[Trash80toG4]

@luRaichu jumped into the project with both feet. He did the board layout, ordered the PCBs, the components, assembled the first one and it's in my grubby little paws for testing! All goes well so far outside of playing whack-a-mole with the gremlins actively undermining my attempts to get a stable testbed up and running.

TwinSlot moniker taken by IIsi card so we're calling this one DoubleBourrage, Double-Stuffer en Francais.



Craziness aside, got the first full test together and it passed with flying colors. Random USB card shows up in TattleTech's PCI Cards report, that's about all I was shooting for at the time. Nice big 1920x1200 main screen. I'd love to see that on a primary workstation with the secondary panel of the TAM active on a KVM setup. But alas, no TAM and no access to its PCI Riser for measurement and adaptation of the board design. Little help there?



Need to get OS9.2 set up on another Compact Flash Card before I try to get much further along. Once the Crescendo/L2 is up and running I have all manner of I/O cards destined for the top, set back Slot_B.

Worse case scenario will be one of my trio of Sonnet TEMPOtrio cards, each with two channels of ATA-133, USB and Firewire. They're about the largest cards I can imagine fitting in the top slot. Tested for fit as it's been the aim all along to use one. It PASSED!. Supposedly works in Alchemy architecture, but they dropped support for Gazelle. But PCI Bridge hacks over at the MLA may prove differently?



It fits! But of course I dropped the friggin' CD and killed it. Need to find another one for the OS9 Lives installer.



The trio is a full on seven inch PCI Card and I snaked it in there sweet as can be. If there are any problems with the MedusaMess and SCSI Optical I can just run an ATAPI Optical off the trio with a DOM on the other ATA133 Channel.

Set up my series of cardboard prototypes to just barely squeeze the trio in far enough for short, RA USB & FW extension cables to be snaked out the vestigial PDS Slot opening.





I set Slot_B back another tenth inch to make just a tad more room for the RA cable breakout and it still fits under the CD mess! I have quite a few other I/O cards to test. So it's gonna be a slog before official rollout. It's been a very long time coming, but product release is nigh!

 

[joevt]

Supposedly works in Alchemy architecture, but they dropped support for Gazelle. But PCI Bridge hacks over at the MLA may prove differently?

Power Mac 6360 doesn't have built-in ATI graphics so it might not have the same issue as Power Mac 6500/TAM (Gazelle).
The solution was a workaround. I would like to actually find the real bug one day.
https://68kmla.org/bb/threads/the-great-gazelle-pci-hack-thread-part-2.38360/post-467825

Does your Power Mac 6360 use the same ROM as the Power Mac 6500/TAM? Get the 64-bit checksum to be sure.
https://tinkerdifferent.com/threads/apple-network-server-macos-based-roms-found.4756/post-41390

Are you using built-in graphics? Tried any GPUs yet?

If you run into any PCI issues, try lspci for Open Firmware to see if the PCI cards are reachable.
https://tinkerdifferent.com/threads/apple-network-server-macos-based-roms-found.4756/post-42356

 

[Trash80toG4]

Small screen's the 6360's built-in graphics set at 640x480 so TT Reports show up nice and big.
Graphics Card in standard slot A1 is driving the big panel at 1920x1200.
ATI Graphics are nice, but take up a PCI address, so Gazelle can handle one less PCI Card than Alchemy?

I imagine 6360/6400/5400 Alchemy Macs all use the same ROM and it's older than that of the Gazelle machines?

 

[joevt]

ATI Graphics are nice, but take up a PCI address, so Gazelle can handle one less PCI Card than Alchemy?

It depends.
Alchemy has a couple options (slot, interrupt, device number):

AAPL,e407 (Power Mac 5400)
slot A1     0x17 @D
slot B1     0x19 @E
slot C1     0x1C @F
slot E1     0x16 @11 // Ethernet? // same interrupt as bandit // seen in 5400 dump

AAPL,Hooper
slot B1     0x1C @E
slot C1     0x1A @F
slot E1     0x18 @11
slot G1     0x16 @13

What model does your Power Mac 6360 have? AAPL,e407 or AAPL,Hooper ?

What device numbers does your riser provide in the Power Mac 6360? Connect a couple PCI cards and capture output from dump-device-tree to find out. We can parse the device numbers from the reg properties.

Gazelle ROM has these options:

AAPL,e407 (Power Mac 5400)
slot A1     0x17 @D
slot B1     0x19 @E
slot C1     0x1C @F
slot E1     0x16 @11

AAPL,e411 (Power Mac 6500)
slot F1     0x18 @12 // ATI graphics
slot A1     0x17 @D // PCI slot
slot B1     0x19 @E // PCI slot riser 2nd slot
slot C1     0x1C @F // PCI slot riser 3rd slot? but no grant/request for bus master?
slot E1     0x16 @11 // COMM slot

AAPL,3400/2400
slot A1     0x1B @D
slot B1     0x1C @E
slot C1     0x1A @F
slot E1     0x18 @11
slot F1     0x19 @12
slot G1     0x16 @13
slot H1     0x17 @14

AAPL,Jumanji
slot A1     0x17 @D
slot B1     0x18 @E
slot C1     0x19 @F
slot E1     0x1A @11
slot F1     0x1B @12
slot G1     0x1C @13
slot H1     0x1C @14

AAPL,Zanzibar
slot E1     0x16 @11
slot A1     0x17 @D
slot B1     0x18 @E
slot C1     0x19 @F

It looks like the Power Mac 6500 adds a separate PCI slot (slot F1) for the built-in ATI graphics. The Gazelle ROM might be applicable to some Alchemy Macs.

If a ROM doesn't probe a PCI slot and you want to add more PCI slots, then a nvramrc patch can be used to probe more slots. We can assign the same interrupt to multiple slots if there's no unused interrupt to connect to.
Heathrow has 6 PCI request/grant pairs (1 high priority grant and 5 round robin grants) and an external arbiter is possible. OHare has maybe one less request/grant pair?

I imagine 6360/6400/5400 Alchemy Macs all use the same ROM and it's older than that of the Gazelle machines?

Right. If it's Alchemy then it more likely uses 6400 ROM. Get the 64-bit checksum or Dump it to be sure.
Power Mac 5400/6400 has Open Firmware 2.0 (different than Copland D11E4 and ANS)
Power Book 3400c, G3 Kanga, Wallstreet, Wallstreet PDQ has Open Firmware 2.0.1.
Power Mac 4400/7200 has Open Firmware 2.0.2.
Power Mac 5500/6500/TAM has Open Firmware 2.0.3.

 

[Trash80toG4]

Thanks, very cool info about available slots, especially Slot_F1 for ATI in Gazelle. I was under the impression that archicture wouldn't support the Mythical Slot C so it's good to see it does! Next it's on to spelunking for Slot_F1 it seems. Not really expecting to find it supported in the connections on the Riser Slot, but if you can identify the Address line for resistive connection we'd be halfway there.

I'm not expecting any problems at all as it's an exact reproduction of the 6x00 riser. It's scrunched way down to fit in the PCI Drawer Macs with Slot_B1 set back for jumper cable extension to the outside world.

Hoping to create a QuadSlot Riser for the Testbed MiniTower some day. As tested, Slot_E1 connections are jumpered from a dummy CSII card to make room for a Gigabit NIC.

 

[joevt]

but if you can identify the Address line for resistive connection we'd be halfway there.

Not sure what this means.
Since slot F1 has device number @12 (hex), I would expect that address line 18 (decimal) is used for the IDSEL signal for that PCI device. That's what the Power Mac 6500 schematics show (sheet 16 for the ATI-264GT device, area A6 in that diagram).
Slot C1 has device number @F so it must use address line 15 for IDSEL.
The PCI Slot in the 6500 gets 3 interrupt signals but I only see two grant/request pairs so the third slot can't be a bus master?

 

[Trash80toG4]

Dunno, all you do is well above my pay grade. I just reverse engineered the 6x00 riser and applied what I'd learned to set up the "Mythical Slot_C" @trag had as the third PCI Slot on one of his clone risers along with a CSII Slot that was taped over.



Apple tied all four interrupts together and came up with their own way of doing PCI, using pins unassigned in the spec . . . Apple!

Everything is in this AI8 doodle, save for the two capacitor connections we discovered when I took a hacksaw to the connector on a 6x00 riser. Intended to buzz it out . . . but Neolithic Hacker Mode kicked in . . .



I do electron plumbing, in what I use for a brain it's all pipes. valves, connector fittings of every kind, gate valves, backflow prevention and whatnot. Kinda works for me. Grew up looking at the LSI and VSLI textbooks as artwork when dad was a boffin at IBM's Glendale Labs and had that stuff laying around.

This is my test rig, it stopped working reliably, but I can bring it back to life if you can identify what might go where to activate Slot_F1 and I can test for it in jumper wire playtime.





I ignored the capacitor connections and everything checked out in a basic fashion using TattleTech's PCI Slot Report. I'd imagine they're there for a very good reason when it comes to full card function. No Clue there, but I can update the artwork.


edit: first thing to do is check to see if there are any "unassigned" pins remaining on the riser connector.

 

[Trash80toG4]

I need to get on with seriously testing DoubleBourrage. Yesterday was my first day back to work after a six month hiatus on workman's comp. Hated forced "retirement" especially the first three months of recovery. But I'm back to limited playtime again. Double edged sword that.

Tangent _______________________________________________________________

If there are three unassigned pins remaining on the "Riser Connector. It's "nominally" Slot_A1++ so there's a chance that Slot_F1 can be activated on the riser. If not, you may be able to find the necessary signals/pin legs in the schematic? In that case, three bodge wires to the board components should do.

Four Slots using the jumpered CSII signals for Gigabit NIC will be very cool. Hrmmm, I wonder if those three lines missing signals are implemented on CSII?

 

[Trash80toG4]

YAY! You got me thinking outside where I'd boxed myself into only using Slot B_1 for USB/FW!

I can just as well throw a second VidCard in there for a three screen Alchemy/Gazelle machine as my test case for TwinSlot!