Three Slot Riser for 6400 - In search of the Mythical Slot C

[Trash80toHP_Mini]

Interesting stuff, been trying to wrap my head around this. Tracked down the original 1996 version of DPCIaDfPMC covering Alchemy (and other first gen. PCI Architectures) only I hope. It's unpolluted by Gazelle (hopefully) and New World complications introduced in the 1997 Second and 1999 revised editions.

1996 - ISBN 13-107236-6
1997 - ISBN -13-107236-6
1999 - ISBN 0-201-52404-X

edit: link to Second Edition is best I could find within edit window.

 

[Trash80toHP_Mini]

If interrupt lines are tied together on the logic board, how does this compute?

 

[joevt]

If interrupt lines are tied together on the logic board, how does this compute?

"combined by OR" means "tied together". What's the issue?

 

[cheesestraws]

If interrupt lines are tied together on the logic board, how does this compute?

Under the right electrical circumstances, just tying signals together does the same thing as an OR gate. It's called a 'wired or'. See https://en-support.renesas.com/knowledgeBase/16978550 - I know little about PCI but this is presumably one of those circumstances.

 

[joevt]

1996 - ISBN 13-107236-6
1997 - ISBN -13-107236-6
1999 - ISBN 0-201-52404-X

edit: link to Second Edition is best I could find within edit window.

ISBN 0-13-107236-6 is for the "Writing FCode Programs" book.
Your 2nd link was for the same 1999 revision as the 3rd link, not 1997.
There's a 1995 version in the PCI Driver Development Kit 2.0 on the Dec 95 Reference Library edition of the Apple Developer CD.

 

[Trash80toHP_Mini]

Sweet! I have that CD on the shelf above my workstation! Title looks it to be well above my pay grade, though.

Yep, searched ISBN in the '99 Revision and jumped to conclusions. Interesting that the PDF versions don't seem to have the ISBN listed?

 

[Trash80toHP_Mini]

I know little about PCI but this is presumably one of those circumstances.

Join the club, thanks for the info! Now I have even more reason to be confused, but you've given me a hand hold to grab onto for the ride.

WAG would be that Slot_B, INTC being wired to Reserved_A9 switches the flip somehow?

 

[Trash80toHP_Mini]

@Daniël here's the latest, slimmed down version for verification of my Slot_B schematic and a first try at the hunt for Slot_C.


Did I get your markup info in correctly?

I'm curious about the four reserved pins for 3.3V magic on a standard PCI Slot. Wondering if apple implemented something on them or if they really are an n.c. situation? If so, they can probably be hijacked for sending your borked ATI signals up a custom riser to implement a Slot?

Comments anyone?

 

[Trash80toHP_Mini]

Made some headway on the Cable Riser testbed reboot. VidCard tested as Slot A1 after every single modification. All Interrupt lines bundled on gray breadboard cable.Next will be doing same for the rest of the lines for mix and match. Address lines bared for resistive hookup.

Think the reboot was a good idea?

 

[Trash80toHP_Mini]

Started to develop patch map at work, finished today. Final check and soldering the jumper wires tomorrow!



Oopsie, forgot to wire the resistor into the Init Device Select jumper.

 

[Trash80toHP_Mini]

Much too tired for a second post:

TwiSlot PCI Riser and ComSlot2 NIC repro in widened CS riser form factor

The schematic for J8 of the 6500 says the reserved pins are for a PCI adapter card. I suppose the PCI adapter card is an Apple specific adapter so it can do weird stuff like requiring reserved pins to handle separate interrupts. The 6x00 two slot Riser is indeed specific to Alchemy/Gazelle and...

tinkerdifferent.com

 

[Daniël]

Excellent! I actually have a three PCI card riser wired up for Slot A, B and C usage, but as of right now I cannot test it, because my PCI extension module is not long enough for the riser to properly clear the case.

Good to know that it works, though

@joevt I saw some questions on the other forum with regards to some of the signals used for Slot C.
Basically, there is a dedicated interrupt for "PCI Slot 3"/Slot C, which is left wired up to the PCI slot (and for some reason, the L2 Cache slot), but the REQ and GNT signals were repurposed for the ATi chip on the 6500 (hence the pins carrying those lines on the 6400 are marked as reserved again on the 6500's PCI slot).

I actually deduced this from comparing a 6400 and 6500 board with a multimeter, tracing where the ATi's REQ and GNT signals come from, which is how I found that those reserved pins indeed have those signals there on the 6400, rather than on the onboard video (as the 6400's Valkyrie-AR isn't a PCI device).
My modified 6500 board with the ATi chip pulled, has this wired as it is on the 6400, to restore Slot C capability.

Though it remains interesting that the O'Hare has seemingly more to offer than is given on Alchemy and Gazelle, as Tanzania can handle up to six PCI devices (clones like the Motorola StarMax 5500 offer five PCI slots, next to onboard ATi PCI graphics).
Tanzania does use a BGA variant of O'Hare, so perhaps it has more pins to expose more REQ/GNT signal pairs.

That, or they're simply not listed in the Gazelle schematics, or at least not obviously.

 

[Trash80toHP_Mini]

Excellent! I actually have a three PCI card riser wired up for Slot A, B and C usage . . .

Thanks! Rewiring on of my collection of three slot risers as you have is probably going to be my next step. The DEC mid-tower recased 6x00 testbed has plenty of headroom. I only need to remove the backplane cover plate from the Logic Boar.. Don't need bracket support as the board is mounted vertically.

I'll be retiring the CableMonster Test Riser.

 

[Trash80toHP_Mini]

Excellent! I actually have a three PCI card riser wired up for Slot A, B and C usage, but as of right now I cannot test it, because my PCI extension module is not long enough for the riser to properly clear the case.

I just ordered up one of these, it might help in your use case:



It'll sit my three slot PC riser with cards vertical, the three slot riser easily clearing the fan on the CPU and the Battery.

PCI RISER-1U-RIGHT Inward Angle Rackmount Adapter Extender Card | eBay

32 Bit Direct Plug In 1U Riser Card. Right Angle (Inward facing). Browse our featured categories. Browse All categories. Although quite a number of manufacturer have adopted the 100-240v (autovolt) power, still a large number of electronics devices may still be mainly 110v.

www.ebay.com

 

[Trash80toHP_Mini]

@Daniël Could you post the wiring diagram for your conversion of that three slot riser, please? Getting ready to buzz mine out and comparing notes seems a good idea.

Thankfully my riser has .01 headers for wire wrapping, planning to bodge Apple's reserved line connections to the adapter card above. Hopefully I won't need to mess with the standard three slot riser?

WAG is that one of the headers for each slot will be Init Device Select. If so, did you wire the resistor Apple employed? Figuring such would be handled by a PC logic board as in the case of Slot_A1 on Apple risers?

 

[Daniël]

I don't really have a wiring harness, I just wired the right pins together based on my earlier findings.
In this case, Slot A is the middle one, and I wired the interrupt jumper pins together for B and C, and then to the right Reserved Pin on the PCI slots:

 

[Trash80toHP_Mini]

Interesting, every Riser I've tried had the bottom slot as a straight thru connection, taking on the ID of whatever board slot in which thet've been placed. This was true across a couple of non-tentacled risers and three or four with tentacles.

The top two slots of mine have the tentacles attached, yours is different?

 

[Daniël]

Yes, you can even see the traces from the lower set of headers go to the lower PCI slot.

 

[Trash80toHP_Mini]

Yep, saw that and also the very odd spacing of the tentacle contacts, thanks for the confirmation. Is that a generic riser or one intended for a specific mfr? The PCB layout seems more complicated than running the bottom slot traces 1:1 would be?

 

[Trash80toHP_Mini]

Here we go! @Daniël do the "PCI Reserved" traces on your riser make it a the way through Slot_A1 to Slot_B1 and Slot_C1? I haven't even bothered to check. I'll be connecting all the jumpers to Slot_A1 as implemented on the successfully tested CableMonster.