Once found in a dark dumpster, I have relentlessly tried to revive this Ultimedia Video Adapter (7-5) for RS/6000:because when I see an adapter ID of 0x86FF on the diag, it *could* be falsely transmitted data on the board coming from the MCU (U15) through U26/27, U32/33, U37/38 to the MIAMI (U43).
http://www.holzapfel.biz/UltimediaVideo7-5.jpg
Here's the tool:
https://www.ardent-tool.com/RS6000/RS6000_7-5.html
I also have the JPEG compression/decompression guest card.
It would be such a nice addition to my -59H or -42T system, and usually when someone claims something being "broken", I usually state they just didn't try hard enough...
It appears to be "broken" in a way that it
a) either falsely reports itself to the bus, like when in RS/6000 diagnostics, the adapter's ID is shown as 0xFFFF or 0x8FFF instead of the actual 0x8F96 ID
b) completely disturbs the Microchannel Bus adapter recognition so that other adapters aren't recognized at all (like Corvette, Audio 7-6, Ethernet 9-K, all show up as ID 0xFFFF in the Diagnostics).
I tried many things to determine what - or where - it is actually broken, like:
- warming the board with a fan, both completely and in different local areas - chilling the board with ice spray, all over and locally
- "baking" the board at ~350 °F / 180 °C like some folks do to revive their GPUs. I know it is not enough to reflow the solder, but success reports pop up here and there
- measuring all MCA Bus pins from the card edge to the MIAMI MCA Interface chip 34G1520, they all beep through. I found one broken via next to U47 and enforced about two dozen by soldering a thin wire through just any suspicious via in that area
- measuring +5 V voltage on the board - of course it's fine
- replacing some of the SMD tantal capacitors near the MCA connection edge
I am close to swapping all generic 74xx logic chips like the 74ABT16245 and 74ABT16652 (U26/27, U32/33, U37/38) which are the *data* bus transceivers between the MIAMI MCA Interface chip and the board-local video processing components on the left side,
But then there's the other error where the card completely hogs the Microchannel bus - so it's not all / only about the data, but possibly also about some MCA control lines. (MIAMI, really? Don't leave me.)address 0x1FFA000C according to MIAMI datasheet.
Digging into the MIAMI's data sheet https://www.ardent-tool.com/tech/miamimca.html
reveals that this is the 2nd iteration of the chip ("Miami Pass 2 (w/o heat spreader) - 34G1520 No Longer in Production") which has been superseded by 34G1521 which HAS a heat spreader, but assumingly remains the same in silicon.
Could it be my MIAMI - lacking the heat spreader - has gone toast?
From the errata list, nothing describes the errors I am seeing.
Looking at the board, some more conditions make me wonder:
- the board has quite some hand made corrections and additions
- some hand written or -corrected PAL / ASIC labels
Those make me assume this might be one of the earlier boards produced, which makes it an interesting piece of history (at least for me).
Looking at the label of U20 it appears to be printed once, but molten off at some point, so maybe U20 got hot one day? It feels just normal in today's "operation" though.
What I found out about the internal structure:
- The MIAMI (U43) acts as a Microchannel Interface Controller that *should* properly handle all control signals on the MCA Bus and translate this board's local address space onto the Microchannel Bus.
- The upper-right part U41, U53 is the AD/DA video frontend
- Left of MIAMI are the bus transceivers (U26/27, U32/33, U37/38) for the local *data* bus between MIAMI and the digital video processor U16 and (probably) microcontroller U15, which *should* be the component delivering the 0x8F96 ID via local data bus
- The mystic PAL logic U20-U25, U28, U30, U34 controls the data bus transceivers' control lines, and probably much more. They are the glue logic for *controlling* everything between the MIAMI and the chips that do the actual video processing.
So what am I missing, what remains there to check?
Unfortunately I don't have a 64 channel logic analyzer to monitor the whole bus - and even then I'm not sure what I would be looking out for.
Any hints?
Thanks,
Christian
Dogrob a 34G1520 off a Passplay or Cheetah?
Christian Holzapfel wrote:
Once found in a dark dumpster, I have relentlessly tried to revive
this Ultimedia Video Adapter (7-5) for RS/6000:
http://www.holzapfel.biz/UltimediaVideo7-5.jpg
Here's the tool:
https://www.ardent-tool.com/RS6000/RS6000_7-5.html
I also have the JPEG compression/decompression guest card.
It would be such a nice addition to my -59H or -42T system, and
usually when someone claims something being "broken", I usually state
they just didn't try hard enough...
It appears to be "broken" in a way that it
a) either falsely reports itself to the bus, like when in RS/6000
diagnostics, the adapter's ID is shown as 0xFFFF or 0x8FFF instead of
the actual 0x8F96 ID
b) completely disturbs the Microchannel Bus adapter recognition so
that other adapters aren't recognized at all (like Corvette, Audio
7-6, Ethernet 9-K, all show up as ID 0xFFFF in the Diagnostics).
I tried many things to determine what - or where - it is actually
broken, like:
- warming the board with a fan, both completely and in different local
areas
- chilling the board with ice spray, all over and locally
- "baking" the board at ~350 °F / 180 °C like some folks do to revive
their GPUs. I know it is not enough to reflow the solder, but success
reports pop up here and there
- measuring all MCA Bus pins from the card edge to the MIAMI MCA
Interface chip 34G1520, they all beep through. I found one broken via
next to U47 and enforced about two dozen by soldering a thin wire
through just any suspicious via in that area
- measuring +5 V voltage on the board - of course it's fine
- replacing some of the SMD tantal capacitors near the MCA connection
edge
I am close to swapping all generic 74xx logic chips like the
74ABT16245 and 74ABT16652 (U26/27, U32/33, U37/38) which are the
*data* bus transceivers between the MIAMI MCA Interface chip and the
board-local video processing components on the left side, because when
I see an adapter ID of 0x86FF on the diag, it *could* be falsely
transmitted data on the board coming from the MCU (U15) through
U26/27, U32/33, U37/38 to the MIAMI (U43).
But then there's the other error where the card completely hogs the
Microchannel bus - so it's not all / only about the data, but possibly
also about some MCA control lines. (MIAMI, really? Don't leave me.)
Digging into the MIAMI's data sheet
https://www.ardent-tool.com/tech/miamimca.html
reveals that this is the 2nd iteration of the chip ("Miami Pass 2 (w/o
heat spreader) - 34G1520 No Longer in Production") which has been
superseded by 34G1521 which HAS a heat spreader, but assumingly
remains the same in silicon.
Could it be my MIAMI - lacking the heat spreader - has gone toast?
From the errata list, nothing describes the errors I am seeing.
Looking at the board, some more conditions make me wonder:
- the board has quite some hand made corrections and additions
- some hand written or -corrected PAL / ASIC labels
Those make me assume this might be one of the earlier boards produced,
which makes it an interesting piece of history (at least for me).
Looking at the label of U20 it appears to be printed once, but molten
off at some point, so maybe U20 got hot one day? It feels just normal
in today's "operation" though.
What I found out about the internal structure:
- The MIAMI (U43) acts as a Microchannel Interface Controller that
*should* properly handle all control signals on the MCA Bus and
translate this board's local address space onto the Microchannel Bus.
- The upper-right part U41, U53 is the AD/DA video frontend
- Left of MIAMI are the bus transceivers (U26/27, U32/33, U37/38) for
the local *data* bus between MIAMI and the digital video processor U16
and (probably) microcontroller U15, which *should* be the component
delivering the 0x8F96 ID via local data bus address 0x1FFA000C
according to MIAMI datasheet.
- The mystic PAL logic U20-U25, U28, U30, U34 controls the data bus
transceivers' control lines, and probably much more. They are the glue
logic for *controlling* everything between the MIAMI and the chips
that do the actual video processing.
So what am I missing, what remains there to check?
Unfortunately I don't have a 64 channel logic analyzer to monitor the
whole bus - and even then I'm not sure what I would be looking out for.
Any hints?
Thanks,
Christian
Look at SSA adapters for a 34G1521, both PCI and MCA use MIAMIs.
Louis Ohland wrote:
Dogrob a 34G1520 off a Passplay or Cheetah?
Christian Holzapfel wrote:
Once found in a dark dumpster, I have relentlessly tried to revive
this Ultimedia Video Adapter (7-5) for RS/6000:
http://www.holzapfel.biz/UltimediaVideo7-5.jpg
Here's the tool:
https://www.ardent-tool.com/RS6000/RS6000_7-5.html
I also have the JPEG compression/decompression guest card.
It would be such a nice addition to my -59H or -42T system, and
usually when someone claims something being "broken", I usually state
they just didn't try hard enough...
It appears to be "broken" in a way that it
a) either falsely reports itself to the bus, like when in RS/6000
diagnostics, the adapter's ID is shown as 0xFFFF or 0x8FFF instead of
the actual 0x8F96 ID
b) completely disturbs the Microchannel Bus adapter recognition so
that other adapters aren't recognized at all (like Corvette, Audio
7-6, Ethernet 9-K, all show up as ID 0xFFFF in the Diagnostics).
I tried many things to determine what - or where - it is actually
broken, like:
- warming the board with a fan, both completely and in different
local areas
- chilling the board with ice spray, all over and locally
- "baking" the board at ~350 °F / 180 °C like some folks do to revive
their GPUs. I know it is not enough to reflow the solder, but success
reports pop up here and there
- measuring all MCA Bus pins from the card edge to the MIAMI MCA
Interface chip 34G1520, they all beep through. I found one broken via
next to U47 and enforced about two dozen by soldering a thin wire
through just any suspicious via in that area
- measuring +5 V voltage on the board - of course it's fine
- replacing some of the SMD tantal capacitors near the MCA connection
edge
I am close to swapping all generic 74xx logic chips like the
74ABT16245 and 74ABT16652 (U26/27, U32/33, U37/38) which are the
*data* bus transceivers between the MIAMI MCA Interface chip and the
board-local video processing components on the left side, because
when I see an adapter ID of 0x86FF on the diag, it *could* be falsely
transmitted data on the board coming from the MCU (U15) through
U26/27, U32/33, U37/38 to the MIAMI (U43).
But then there's the other error where the card completely hogs the
Microchannel bus - so it's not all / only about the data, but
possibly also about some MCA control lines. (MIAMI, really? Don't
leave me.)
Digging into the MIAMI's data sheet
https://www.ardent-tool.com/tech/miamimca.html
reveals that this is the 2nd iteration of the chip ("Miami Pass 2
(w/o heat spreader) - 34G1520 No Longer in Production") which has
been superseded by 34G1521 which HAS a heat spreader, but assumingly
remains the same in silicon.
Could it be my MIAMI - lacking the heat spreader - has gone toast?
From the errata list, nothing describes the errors I am seeing.
Looking at the board, some more conditions make me wonder:
- the board has quite some hand made corrections and additions
- some hand written or -corrected PAL / ASIC labels
Those make me assume this might be one of the earlier boards
produced, which makes it an interesting piece of history (at least
for me).
Looking at the label of U20 it appears to be printed once, but molten
off at some point, so maybe U20 got hot one day? It feels just normal
in today's "operation" though.
What I found out about the internal structure:
- The MIAMI (U43) acts as a Microchannel Interface Controller that
*should* properly handle all control signals on the MCA Bus and
translate this board's local address space onto the Microchannel Bus.
- The upper-right part U41, U53 is the AD/DA video frontend
- Left of MIAMI are the bus transceivers (U26/27, U32/33, U37/38) for
the local *data* bus between MIAMI and the digital video processor
U16 and (probably) microcontroller U15, which *should* be the
component delivering the 0x8F96 ID via local data bus address
0x1FFA000C according to MIAMI datasheet.
- The mystic PAL logic U20-U25, U28, U30, U34 controls the data bus
transceivers' control lines, and probably much more. They are the
glue logic for *controlling* everything between the MIAMI and the
chips that do the actual video processing.
So what am I missing, what remains there to check?
Unfortunately I don't have a 64 channel logic analyzer to monitor the
whole bus - and even then I'm not sure what I would be looking out for.
Any hints?
Thanks,
Christian
https://www.ebay.com/itm/271310985827
34G1521 does not have a heat spreader.
Louis Ohland wrote:
Look at SSA adapters for a 34G1521, both PCI and MCA use MIAMIs.
Louis Ohland wrote:
Dogrob a 34G1520 off a Passplay or Cheetah?
Christian Holzapfel wrote:
Once found in a dark dumpster, I have relentlessly tried to revive
this Ultimedia Video Adapter (7-5) for RS/6000:
http://www.holzapfel.biz/UltimediaVideo7-5.jpg
Here's the tool:
https://www.ardent-tool.com/RS6000/RS6000_7-5.html
I also have the JPEG compression/decompression guest card.
It would be such a nice addition to my -59H or -42T system, and
usually when someone claims something being "broken", I usually
state they just didn't try hard enough...
It appears to be "broken" in a way that it
a) either falsely reports itself to the bus, like when in RS/6000
diagnostics, the adapter's ID is shown as 0xFFFF or 0x8FFF instead
of the actual 0x8F96 ID
b) completely disturbs the Microchannel Bus adapter recognition so
that other adapters aren't recognized at all (like Corvette, Audio
7-6, Ethernet 9-K, all show up as ID 0xFFFF in the Diagnostics).
I tried many things to determine what - or where - it is actually
broken, like:
- warming the board with a fan, both completely and in different
local areas
- chilling the board with ice spray, all over and locally
- "baking" the board at ~350 °F / 180 °C like some folks do to
revive their GPUs. I know it is not enough to reflow the solder, but
success reports pop up here and there
- measuring all MCA Bus pins from the card edge to the MIAMI MCA
Interface chip 34G1520, they all beep through. I found one broken
via next to U47 and enforced about two dozen by soldering a thin
wire through just any suspicious via in that area
- measuring +5 V voltage on the board - of course it's fine
- replacing some of the SMD tantal capacitors near the MCA
connection edge
I am close to swapping all generic 74xx logic chips like the
74ABT16245 and 74ABT16652 (U26/27, U32/33, U37/38) which are the
*data* bus transceivers between the MIAMI MCA Interface chip and the
board-local video processing components on the left side, because
when I see an adapter ID of 0x86FF on the diag, it *could* be
falsely transmitted data on the board coming from the MCU (U15)
through U26/27, U32/33, U37/38 to the MIAMI (U43).
But then there's the other error where the card completely hogs the
Microchannel bus - so it's not all / only about the data, but
possibly also about some MCA control lines. (MIAMI, really? Don't
leave me.)
Digging into the MIAMI's data sheet
https://www.ardent-tool.com/tech/miamimca.html
reveals that this is the 2nd iteration of the chip ("Miami Pass 2
(w/o heat spreader) - 34G1520 No Longer in Production") which has
been superseded by 34G1521 which HAS a heat spreader, but assumingly
remains the same in silicon.
Could it be my MIAMI - lacking the heat spreader - has gone toast?
From the errata list, nothing describes the errors I am seeing.
Looking at the board, some more conditions make me wonder:
- the board has quite some hand made corrections and additions
- some hand written or -corrected PAL / ASIC labels
Those make me assume this might be one of the earlier boards
produced, which makes it an interesting piece of history (at least
for me).
Looking at the label of U20 it appears to be printed once, but
molten off at some point, so maybe U20 got hot one day? It feels
just normal in today's "operation" though.
What I found out about the internal structure:
- The MIAMI (U43) acts as a Microchannel Interface Controller that
*should* properly handle all control signals on the MCA Bus and
translate this board's local address space onto the Microchannel Bus.
- The upper-right part U41, U53 is the AD/DA video frontend
- Left of MIAMI are the bus transceivers (U26/27, U32/33, U37/38)
for the local *data* bus between MIAMI and the digital video
processor U16 and (probably) microcontroller U15, which *should* be
the component delivering the 0x8F96 ID via local data bus address
0x1FFA000C according to MIAMI datasheet.
- The mystic PAL logic U20-U25, U28, U30, U34 controls the data bus
transceivers' control lines, and probably much more. They are the
glue logic for *controlling* everything between the MIAMI and the
chips that do the actual video processing.
So what am I missing, what remains there to check?
Unfortunately I don't have a 64 channel logic analyzer to monitor
the whole bus - and even then I'm not sure what I would be looking
out for.
Any hints?
Thanks,
Christian
Once found in a dark dumpster, I have relentlessly tried to revive this Ultimedia Video Adapter (7-5) for RS/6000
https://www.ebay.com/itm/271310985827
34G1521 does not have a heat spreader.
You can see what appears to be uniformly spaced gouges from the right, seemingly touching the edge of A2.
Also, looking at the photo, the A1 pin on the edge connector looks
really suspicious. This is the -CD SETUP line, responsible for putting
the adapter into the setup mode. And that's when the adapter ID is
obtained, among other things. So, an unreliable connection could
definitely cause the described behavior.
Before you dive into that, it may be worth checking if the oscillatorsare stable and outputting what they are supposed to.
I agree that the issue most likely lies somewhere between the MCA edge connector and the MIAMI interface chip (hopefully not inside the ASIC itself).
You could also test the adapter in another machine (even a PS/2) and
check if the behavior changes in any way.
Look at SSA adapters for a 34G1521, both PCI and MCA use MIAMIs.
You can see what appears to be uniformly spaced gouges from the right, seemingly touching the edge of A2.
I wonder what the solder side looks like...
I once checked in my -59H, and it didn't work. Don't remember the precise error pattern though. Will repeat that step.
Looks like many PALs/GALs are present. Please keep in mind, that all of these have a limited retention.I once checked in my -59H, and it didn't work. Don't remember the precise error pattern though. Will repeat that step.Same thing: No adapters recognized.
Since the -59H doesn't have in-ROM diagnostics and it can't IPL from the Corvette, the last chance of seeing anything would be booting diags from the CD-ROM on the internal SCSI port - but I've seen enough.
May try the P70 some time, but I bet the Ultimedia would disable the on-planar MCA DBA controller as well, leaving not more than booting from the RefDisk...
If it is not essential, what else might be broken too?
What could be wrong here, technically?
Happy New Year everyone!IBM miss packaging some files?
I finally progressed. After more probing of the MCA control lines I found U45, U48 and U49 (74F00, 74F245 and 74F10) dealing with the CD_CHRDY, /S0, /S1, /BURST, /CMD and /ADL signals.
The 74F245 had the most pins and complexity, so I swapped it first against another that I ripped off a dead Token RIng adapter.
And surprise: The system now boots from disk into AIX 4.3.3, identifies the adapter as "Ultimedia Video I/O Adapter sr0", assigns resources to the card and shows it in the diagnostics. Weeee!!
However, functional tests fail:
- UMS applications complain about "Video adapter not found" altough seen in the device list
- Device specific diagnostics fail reporting "Cannot run /usr/lpp/diagnostics/da/dsun". That file does not exist, and does not seem to come with the 8F96 driver package. It is not contained in the PCI Ultimedia Video Adapter device drivers either. Did
http://www.holzapfel.biz/8F96/8F96_DiagnosticsNotFound.jpgsort of discrete EEPROM, Flash or others.
- Viewing the Vital Product Data (VPD) in the Diagnostics reveals all empty fields, opposed to the DFE5 Ultimedia Audio Adapter:
http://www.holzapfel.biz/8F96/8F96_DisplayVPD.jpg
http://www.holzapfel.biz/8F96/DFE5_DisplayVPD.jpg
- Diagnostics allow me setting some VPD fields, however, opposed to the DFE5, all fields are read-write, which may not be intended:
http://www.holzapfel.biz/8F96/8F96_AlterVPD.jpg
http://www.holzapfel.biz/8F96/DFE5_AlterVPD.jpg
- On the good end, I committed some random data to the "USER" field, and it persisted through a power cycle as expected.
So if the initial VPD is missing, where on the board would it be stored? MIAMI's datasheet doesn't reveal much, so I would assume it to be part of the microcontroller U15. It needs to be a component with non-volatile memory, but I couldn't identify any
Is the VPD data essential for the adapter's functionality?
How did it go lost?
Would setting the data to proper values help?
What are the proper values?
If it is not essential, what else might be broken too?
What could be wrong here, technically?
Nice work, Christian!
The complex and planar VPD info isn't really "vital" in the case of the
late PS/2s. But I have no clue how pedantic AIX might be about this stuff.
However, functional tests fail:
- UMS applications complain about "Video adapter not found" altough seen in the device list
Without proper documentation on the vendor defined use of those fields, it looks pretty useless...
Following Louis' path into the ODM database by
odmget -q "name = sr0" CuVPD"
reveals that all information I entered seems to be stored inside the ODM - at least.
"When bits 6 and 7 in POS Register 5 are set to 0, reading POS
Registers 6 and 7 will return the channel-check-status information.
This information can be status or a pointer to the status."
[..]
Without proper documentation on the vendor defined use of those fields, it looks pretty useless...
The MIAMI spec delivers.
Before you dive into that, it may be worth checking if the oscillators
are stable and outputting what they are supposed to.
It's about the journey, right?
Sysop: | Keyop |
---|---|
Location: | Huddersfield, West Yorkshire, UK |
Users: | 307 |
Nodes: | 16 (2 / 14) |
Uptime: | 66:23:24 |
Calls: | 6,915 |
Files: | 12,379 |
Messages: | 5,431,756 |