BlueSCSI on Macintosh Portable

[Sideburn]

It also depends on whether people want an update.

Based on all my investigation, there's nothing wrong with the RAM card. It's performing exactly as it should.

All these side effects of increased overhead due to memory management and the bug in CPU wait states are from things that are not necessarily under my direct control.

Yeah I follow. And I’m sure most people don’t even notice. I didn’t. The only reason I noticed was because the Talking Moose was getting all distorted and talking super slow after sleep.

buts it is quite a difference in speed drop. Cuts it in half almost.

Especially after sleep. And it makes system 7 pretty sluggish overall.

I’ll take less ram over loss of performance any day. One more meg would be enough even.

 

[Sideburn]

Ideally a switch on the card for performance vs. memory would be cool. Maybe a v2.0…

 

[alxlab]

https://www.tech-insider.org/mac/research/acrobat/9204.pdf

It's maybe "intentionally" designed to behave the way it dees right now based on the info above. Perhaps it's spending extra time always checking for an installed PDS card when the total ram is above 5MB on a Macintosh Portable and above 4MB on the backlit Macintosh Portable.

The performance drop is a bit sad but dems the breaks I guess.

 

[Sideburn]

Would it be possible to put a dip switch on the board so the user can select the amount of RAM they want?

 

[Androda]

I have now worked through the scenarios and these are my discoveries on the M5120 Portable:
* Upgrade card set to 4MB or less: No perf degradation (no degradation after sleep)
* Upgrade card set to 5MB or more: Perf degradation (worse after sleep)

So it looks like for some reason or another, RAM-slot upgrade cards are penalized when above these arbitrary thresholds. It's unclear why.

 

[alxlab]

I have now worked through the scenarios and these are my discoveries on the M5120 Portable:
* Upgrade card set to 4MB or less: No perf degradation
* Upgrade card set to 5MB or more: Perf degradation

So it looks like for some reason or another, RAM-slot upgrade cards are penalized when above these arbitrary thresholds. It's unclear why.

If it's related to the PDS slot mem management/limitation I'd expect degradation of performance past a 3MB ram card on a backlit Macintosh Portable.

***If the PDS slot is used for other peripherals, then the maximum amount of RAM (by using a RAM expansion card) is 5
MB for the Macintosh Portable, and 4 MB for the backlit Macintosh Portable.

It'd be kinda weird since I don't think any of us actually have a PDS card installed. Wonder if those old PDS ram expansions had the same performance issue.

 

[Androda]

If it's related to the PDS slot mem management/limitation I'd expect degradation of performance past a 3MB ram card on a backlit Macintosh Portable.

It'd be kinda weird since I don't think any of us actually have a PDS card installed. Wonder if those old PDS ram expansions had the same performance issue.

PDS probably doesn't have the perf penalty, because a PDS expansion card would be able to trigger EXT.DTACK independently of the glue logic.

 

[SuperSVGA]

So from what I understand /DTACK is different above the first 5MB of address space, I think it has more wait states due to longer timing on peripheral addresses. This shouldn't be that much of a problem though, because the the problem would appear on first boot and not just sleep/wake.

You can try pulling /EXT.DTACK low for just moment along with your /CS to get /DTACK immediately.

 

[Sideburn]

So from what I understand /DTACK is different above the first 5MB of address space, I think it has more wait states due to longer timing on peripheral addresses. This shouldn't be that much of a problem though, because the the problem would appear on first boot and not just sleep/wake.

You can try pulling /EXT.DTACK low for just moment along with your /CS to get /DTACK immediately.

The problem does appear on first boot (cuts cpu speed in half with 7mb). Then after sleep/wake it cuts that in half again roughly.

 

[alxlab]

The problem does appear on first boot (cuts cpu speed in half with 7mb). Then after sleep/wake it cuts that in half again roughly.

Yeah was just reading the Macintosh Portable can throttle CPU speed. Wonder if disabling automatic sleep would do anything.

The Portable has the ability to slow the microprocessor
from its normal speed of approximately 16 MHz to
about 1 MHz during idle periods. This feature is called
system rest and is used to reduce power consumption
by the 68HCOOO. The computer will switch to system
rest if no input device activity occurs for fifteen
seconds. The system returns to full speed when a user
presses a key, moves the trackball or mouse, or a
peripheral requests attention.
System rest may interfere with the operation of some
programs. If you need to disable system rest, hold
down the <Option> key and click anywhere on the
words Minutes Until Automatic Sleep. The dialog box
appears in Figure II-B. Then click Don't Rest and OK

Reference: https://mirrors.apple2.org.za/ftp.apple.asimov.net/documentation/macintosh/atp_macintoshportable.pdf

 

[Sideburn]

Yeah was just reading the Macintosh Portable can throttle CPU speed. Wonder if disabling automatic sleep would do anything.



Reference: https://mirrors.apple2.org.za/ftp.apple.asimov.net/documentation/macintosh/atp_macintoshportable.pdf

I went through all that meticulously when I was trying to figure out what was going on before I discovered it was the ram expansion. It made no difference. I’ve had rest off and sleep set to max and off when plugged in.

 

[Androda]

I have started work on an alternate firmware and switch-mod-PCB for the RAM card. This will allow switching between the following modes:
* M5120 "Low RAM" (4MB on expansion card)
* M5120 "High RAM" (7MB on expansion card)
* M5126 "Low RAM" (3MB on expansion card)
* M5126 "High RAM" (7 MB on expansion card)

This will allow people to choose between CPU performance and RAM expansion (but not while the system is powered on - that would be bad).

 

[Sideburn]

I have started work on an alternate firmware and switch-mod-PCB for the RAM card. This will allow switching between the following modes:
* M5120 "Low RAM" (4MB on expansion card)
* M5120 "High RAM" (7MB on expansion card)
* M5126 "Low RAM" (3MB on expansion card)
* M5126 "High RAM" (7 MB on expansion card)

This will allow people to choose between CPU performance and RAM expansion (but not while the system is powered on - that would be bad).

Thats sounds great!

 

[Sideburn]

Can you salvage parts of the current boards if we send them back to you?

 

[Androda]

can you salvage parts of the current boards if we send them back to you?

This design is a "modboard", so it would be glued to the main PCB and then it's just four jumper wires. There's no need to order new "full PCBs" since I can just reprogram the CPLD and use a few more of its input pins as configuration.

 

[Sideburn]

After thats done you can make a replacement floppy cable with a connector to power the blueSCSI on wake from sleep to fix that delay

 

[Sideburn]

This design is a "modboard", so it would be glued to the main PCB and then it's just four jumper wires. There's no need to order new "full PCBs" since I can just reprogram the CPLD and use a few more of its input pins as configuration.

Nice! thats even better.

 

[Androda]

Nice! thats even better.

Yeah, no need to rework the whole PCB when there are only a few pins that need tweaking. Future revisions of the RAM card (if they ever sell out) will have the switches by default.

 

[Sideburn]

Yeah, no need to rework the whole PCB when there are only a few pins that need tweaking. Future revisions of the RAM card (if they ever sell out) will have the switches by default.

And then you when you get rich and famous, the Original boards will be worth more.

 

[SuperSVGA]

Ah, so it's the cpu glue which controls that. Is there a register or address reference available?

//---------------------------------------------------
// HcMac Clock Change Addresses
//---------------------------------------------------

#define Clock16M                 $FE0000                 // Go to sixteen megahertz clock throughput
#define Clock1M                  $FE0002                 // Go to one megahertz clock throughput (adds 64 wait states)

Not sure if there's more undocumented speeds or not...