http://ps-2.kev009.com/ohland/Spock_Bermuda/8556_SCSI.html

NOTE: Folks, for years, I was unaware IF the on-board SCSI of the 8556 /
9576 (Bermuda) were actually local bus -OR- if they were attached to the
Micro Channel. They -ARE- local bus.

"In prior systems having storage controllers, the controllers typically comprised a group of disconnected state machines which are controlled by
the system microprocessor. In such systems, the "intelligence' which
runs the system, as to SCSI transfers, is the system microprocessor,
with the SCSI controller (or controller chip set) being unaware of any
pending commands or even the next phase of SCSI transfer which is to
occur. While the system microprocessor is truly in control of the
transfers, it is also very busy with other system operations and very
slow in comparison to hardware logic speeds. This results in performance degradation of the SCSI subsystem."

There is more stuff involving the access by the SCSI Controller to the
local bus, it isn't attached to the Micro Channel bus, but it does
arbitrate. Somehow.

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

On 16.04.2023 4:59, Louis Ohland wrote:

http://ps-2.kev009.com/ohland/Spock_Bermuda/8556_SCSI.html

NOTE: Folks, for years, I was unaware IF the on-board SCSI of the 8556 /
9576 (Bermuda) were actually local bus -OR- if they were attached to the Micro Channel. They -ARE- local bus.

Yep, we have verified this some time ago already. It was as simple as
measuring continuity between the CPU pins and the SCSI bus controller
chip. The 8556/7 and 9556/7 pages already have it marked accordingly.

Bermuda (9576/7 non i/s) however uses an ordinary multi-chip
MCA-attached SCSI subsystem.

I wonder what the difference is between the 64F4760 SCSI controller used
on the 8556/7 planars and the 91F9906 one from 9556/7.

The chips were not listed in the ASIC/GA list. I have fixed that meow: https://www.ardent-tool.com/tech/ASIC_List.html#Storage

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

But never explained.

Tomas Slavotinek wrote:

Yep, we have verified this some time ago already. It was as simple as measuring continuity between the CPU pins and the SCSI bus controller
chip. The 8556/7 and 9556/7 pages already have it marked accordingly.

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

Tom, my SWAG is the earlier SCSI chipsets were unable to handle the
local bus traffic. Or it could be the earlier system chipsets needed
better timing windows or whatnot.

It may be that the Bermuda was the pinnacle of the IBM MCA desktop
systems where the system chipset was capable enough to use the Spock
chipset "as-is"

https://patents.google.com/patent/US5371861A/en

Note that the patent sez "While here illustrated as being a device
connected directly with the local processor bus 34, the present
invention also contemplates that the SCSI controller 40 may be provided
as an option device connected through the I/ O bus 44 mentioned hereinafter"

The BIC in the 8556/9556 is?

Tomas Slavotinek wrote:

On 16.04.2023 4:59, Louis Ohland wrote:

http://ps-2.kev009.com/ohland/Spock_Bermuda/8556_SCSI.html

NOTE: Folks, for years, I was unaware IF the on-board SCSI of the 8556
/ 9576 (Bermuda) were actually local bus -OR- if they were attached to
the Micro Channel. They -ARE- local bus.

Yep, we have verified this some time ago already. It was as simple as measuring continuity between the CPU pins and the SCSI bus controller
chip. The 8556/7 and 9556/7 pages already have it marked accordingly.

Bermuda (9576/7 non i/s) however uses an ordinary multi-chip
MCA-attached SCSI subsystem.

I wonder what the difference is between the 64F4760 SCSI controller used
on the 8556/7 planars and the 91F9906 one from 9556/7.

The chips were not listed in the ASIC/GA list. I have fixed that meow: https://www.ardent-tool.com/tech/ASIC_List.html#Storage

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

Indeed. The patent seems rather rushed (per IBM's standards), but better
than nothing...

On 16.04.2023 11:04, Louis Ohland wrote:

But never explained.

Tomas Slavotinek wrote:

Yep, we have verified this some time ago already. It was as simple as
measuring continuity between the CPU pins and the SCSI bus controller
chip. The 8556/7 and 9556/7 pages already have it marked accordingly.

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

Yes, the traditional IBM SCSI chipsets were designed to interface with
MCA only - different bus would require a different SCSI controller chip
or bus interface chip (Malibu).

Local bus SCSI is cheaper and can perform quite good, thanks to the bus
speed. The patent mentions that the local bus SCSI controller can be
attached to MCA but it's unclear whether this bus is supported by the controller itself or if it would require an additional bridge.

The BIC seems to be integrated to one ASIC together with the memory
controller, etc.:

-8556/7: U58 92F1495
-8556/7: U71 96F7690

On 16.04.2023 11:25, Louis Ohland wrote:

Tom, my SWAG is the earlier SCSI chipsets were unable to handle the
local bus traffic. Or it could be the earlier system chipsets needed
better timing windows or whatnot.

It may be that the Bermuda was the pinnacle of the IBM MCA desktop
systems where the system chipset was capable enough to use the Spock
chipset "as-is"

https://patents.google.com/patent/US5371861A/en

Note that the patent sez "While here illustrated as being a device
connected directly with the local processor bus 34, the present
invention also contemplates that the SCSI controller 40 may be provided
as an option device connected through the I/ O bus 44 mentioned
hereinafter"

The BIC in the 8556/9556 is?

Tomas Slavotinek wrote:

On 16.04.2023 4:59, Louis Ohland wrote:

http://ps-2.kev009.com/ohland/Spock_Bermuda/8556_SCSI.html

NOTE: Folks, for years, I was unaware IF the on-board SCSI of the
8556 / 9576 (Bermuda) were actually local bus -OR- if they were
attached to the Micro Channel. They -ARE- local bus.

Yep, we have verified this some time ago already. It was as simple as
measuring continuity between the CPU pins and the SCSI bus controller
chip. The 8556/7 and 9556/7 pages already have it marked accordingly.

Bermuda (9576/7 non i/s) however uses an ordinary multi-chip
MCA-attached SCSI subsystem.

I wonder what the difference is between the 64F4760 SCSI controller
used on the 8556/7 planars and the 91F9906 one from 9556/7.

The chips were not listed in the ASIC/GA list. I have fixed that meow:
https://www.ardent-tool.com/tech/ASIC_List.html#Storage

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

We have no performance figures to compare the 8556, 9556, and 9576 SCSI implementations. Were the 56 versions a cut-rate solution -OR- was it a
happy co-incidence where the local bus PLUS a SLC flavor made a pretty
good performer that wasn't expensive...

Seems to break the Fast, Cheap, or Good rule...

Odd. A 386SLC or 486SLC system with a frugal SCSI implementation

Tomas Slavotinek wrote:

Yes, the traditional IBM SCSI chipsets were designed to interface with
MCA only - different bus would require a different SCSI controller chip
or bus interface chip (Malibu).

Local bus SCSI is cheaper and can perform quite good, thanks to the bus speed. The patent mentions that the local bus SCSI controller can be
attached to MCA but it's unclear whether this bus is supported by the controller itself or if it would require an additional bridge.

The BIC seems to be integrated to one ASIC together with the memory controller, etc.:

-8556/7: U58 92F1495
-8556/7: U71 96F7690

On 16.04.2023 11:25, Louis Ohland wrote:

Tom, my SWAG is the earlier SCSI chipsets were unable to handle the
local bus traffic. Or it could be the earlier system chipsets needed
better timing windows or whatnot.

It may be that the Bermuda was the pinnacle of the IBM MCA desktop
systems where the system chipset was capable enough to use the Spock
chipset "as-is"

https://patents.google.com/patent/US5371861A/en

Note that the patent sez "While here illustrated as being a device
connected directly with the local processor bus 34, the present
invention also contemplates that the SCSI controller 40 may be
provided as an option device connected through the I/ O bus 44
mentioned hereinafter"

The BIC in the 8556/9556 is?

Tomas Slavotinek wrote:

On 16.04.2023 4:59, Louis Ohland wrote:

http://ps-2.kev009.com/ohland/Spock_Bermuda/8556_SCSI.html

NOTE: Folks, for years, I was unaware IF the on-board SCSI of the
8556 / 9576 (Bermuda) were actually local bus -OR- if they were
attached to the Micro Channel. They -ARE- local bus.

Yep, we have verified this some time ago already. It was as simple as
measuring continuity between the CPU pins and the SCSI bus controller
chip. The 8556/7 and 9556/7 pages already have it marked accordingly.

Bermuda (9576/7 non i/s) however uses an ordinary multi-chip
MCA-attached SCSI subsystem.

I wonder what the difference is between the 64F4760 SCSI controller
used on the 8556/7 planars and the 91F9906 one from 9556/7.

The chips were not listed in the ASIC/GA list. I have fixed that meow:
https://www.ardent-tool.com/tech/ASIC_List.html#Storage

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

http://ps-2.kev009.com/ohland/Spock_Bermuda/8556_SCSI.html

Not much more, some little nuggets on the enabling or disabling of RAM
drivers, but nothing earth-shattering. Might be time to walk off from
this page.

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)