Memory adventure: MicroMac SIMMdoubler II and 68851 MMU in a Macintosh II

[David Cook]

As shipped, the Macintosh II is deceptively on the challenging side of the Macintosh technology curve: It...
1. Had multiple ROM revisions but doesn't include a ROM SIMM slot (like the IIx) to swap in a modern replacement
2. Didn't support 1440K drives
3. Didn't recognize memory SIMMs bigger than 1MB unless the SIMMs included compensating circuitry (such as a GAL)
4. Didn't support 32-bit mode (none of the ROM revisions are 32-bit clean). Software can compensate for this.
5. Didn't include a proper MMU (memory management unit)

The easiest path is to just install eight 1 MB SIMMs and be happy with 8 MB total. If you want more, you'll need to install both a Motorola 68851 MMU and special SIMMs. If you want 32MB or 128MB, you'll also need to upgrade the ROMs (see https://wiki.preterhuman.net/Macintosh_II).

I'll start with a Macintosh II that the seller said would not power up. A while back they had upgraded to a battery holder but believed it must have run out of power. Despite not including a picture of the inside of the machine, there were enough other valuable accessories in the auction to make it worth buying.

I was greeted by this glorious sight. It doesn't seem like an official AppleCare battery holder upgrade. That's a 18650 rechargeable lithium ion cell.


I have so many concerns here. The soldering job, the lack of heat shrink insulation, and the exact opposite choice of wire colors to battery polarity. Mostly though, why only 'fix' one of the batteries?



I don't want to get derailed here from the memory upgrade. It is a simple matter to replace the 18650 and dead cell with on-board battery holders and fresh proper cells. (The HC132 power-on chip was also dead.)

This Macintosh has the stock Apple PMMU. It does not support more than 8MB of memory, nor does it support System 7 virtual memory. If you see a black chip instead of a gold chip, that's the original PMMU.


Use caution when removing the PMMU. More than once, I have pulled pins when removing the PMMU from the socket. Although this seems scary, if you align the pin notches with the oval hole, it press fits back in nicely. That is, the pins appear to be more like connectors than soldered pins.



Although the Apple PMMU is kindly socketed, it seems nearly impossible to find the replacement Motorola 68851 MMU. I found only one. It was being sold for 'gold scrap' and hadn't been handled with love.



I know what you're thinking. Even though none of the pins are broken off, this thing has not exactly been stored in a static bag or protective foam. This isn't going to work, right?

We start with a first-order lever to lift the completely flattened pins. Rocking the dental pick against the chip's ceramic package ensures each pin is lifted with a smooth motion and cannot be bent beyond 90 degrees. In the picture below, note the great progress in lifting the pins without breakage.



Next, we'll use @jwg1962 's mechanical pencil trick to straighten each pin individually. Removing the pencil lead from the tip of a 0.5 mm mechnical pencil creates a nice tube socket with a really long lever arm for precise control. See how straight most of the pins are looking now?



Now that the pins are lifted and straightened, they need to be aligned into rows and columns with respect to each other. I find female socket headers work nicely when the pins are nearly in shape.



Below is the 68851 chip installed. Yes, it works!

Also note the pointless (but fun) upgrade of the FPU to a 68882, the ROMs to the latest revision, a SWIM chip for 1440K disks, and battery holders.



Memory

I've used tall composite memories to max out my IIx to 128MB. I have some 4MB GAL SIMMs on the way. However, something more interesting caught my eye.

There was an auction for a Performa 600 (IIvi) with four MicroMac SIMMdoubler II boards. The SIMMdoubler physically and logically combines two standard SIMMs (1 MB or 4 MB) to a single motherboard SIMM socket. The logical interface is a GAL -- which makes 4 MB SIMMs compatible with the II and IIx.



Amusingly, the fourth SIMMdoubler does not fit vertically in certain machines, like the IIvi. So, MicroMac provides a horizontal adapter. Here it is in action.



Here is the right-angle connector with the other doubler SIMMs removed.



Removing the right-angle adapter from the SIMMdoubler is a chore. Unlike a Macintosh SIMM socket, the right-angle adapter socket has solid brackets on both sides (I guess for strength). There are two little latches that need to be pressed in as you try to wiggle the SIMMdoubler out. Do not attempt this if you are in a bad mood or have been drinking.



Below is the adapter after having been removed. Since I intend to use all four SIMMdoublers in the Mac II, I don't need the right angle adapter so I just threw it away.



Below are the SIMMdoublers in the Mac II. They fit snugly against each other. Notice that the upper memory SIMM sits above the height of the SIMMdoubler circuit board.



To test whether these are too tall, I placed a sheet of paper on top of the SIMMdoublers and then placed the Macintosh II's upper tray in place. Although the screw holes (purple below) showed the tray seemed to resting flat, I was unable to remove the sheet of paper. This indicates that the tray is likely pressing down on the SIMMdoublers and straining the SIMM sockets and motherboard.



With Apple's 32-Bit System Enabler software installed, the computer is successfully able to access more than 8MB. That's great.



Oh wait. 4 MB SIMMs x 2 per board x 4 SIMMdoublers = 32 MB. Why are we only seeing 16 MB? The Macintosh II and IIx only recognize 256KB, 1 MB, 4 MB, or 16 MB SIMMs. The '8 MB' doubled SIMMs are seen as just 4 MB each. 4 x 4 =16 MB. Worst case, I can remove a SIMM from the top slot of each SIMMdoubler to provide enough room for the upper tray to sit flat and still get 16 MB of RAM. I can then add 1 MB x 4 SIMMs to the standard motherboard bank B to get 20 MB total.

Even though the SIMMdoubler does not expressly indicate support for 16 MB SIMMs, I may try it as an experiment. However, given that both the Macintosh II motherboard and the SIMMdoubler SIMM slots all use plastic retention clips rather than metal, I am hesitant to play with them.

- David

P.S. No, I did not actually throw away the SIMMdoubler right-angle connector.

 

[Phipli]

Also note the pointless (but fun) upgrade of the FPU to a 68882

Far from pointless! The 68882 is much faster

Important for Excel power users and ensuring your tax returns are done on time.

 

[David Cook]

Far from pointless! The 68882 is much faster

Important for Excel power users and ensuring your tax returns are done on time.

I just searched for '68881 vs 68882' and I see the subject is a source of flame wars. : )

 

[Phipli]

I just searched for '68881 vs 68882' and I see the subject is a source of flame wars. : )

No idea if it is, but Motorola claimed it was twice as fast.

 

[David Cook]

I have three different types of 30 pin 16 MB SIMMs:
1. 8 chips in two rows mounted horizontally
2. 8 chips mounted vertically
3. Two modern chips mounted horizontally

The first type is too tall to fit in the first SIMM slot of the SIMMdoubler.

I tested the other two by installing four of the same type in the first SIMM slot of the SIMMdoubler. The dip switches were set 'on' 2 & 3. This is the "4 MB in first slot, nothing in second slot" configuration. I tried these in bank A of the Macintosh II. Immediately on power up, I got the Sad Mac chimes. Therefore, I conclude that the SIMMdoubler is not compatible with 16MB SIMMs, or at least not in the Macintosh II in bank A, or the DIP switches need to be set differently. Because the SIMMdoubler and Macintosh II slots all have plastic retention tabs (rather than metal), I was unwilling to try every combination.

So, it looks like the SIMMdoubler is not useful for providing 8 MB or 16 MB SIMMs in the Macintosh II.

However, I did try 4 MB SIMMs in the lower slot with nothing in the top slot. This worked fine. This means you can get 4 x 4MB = 16 MB per bank in the Macintosh II using the SIMMdoubler. And, the top disk drive tray now fits because the top SIMM slot is empty.

- David

 

[LaPorta]

I'm fairly certain they only changed one battery out because the other is for keeping time...and apparently they didn't care outside of the machine starting up.

 

[CircuitBored]

Great work and write-up as usual. It's always a pleasure to read your posts. Fingers crossed that that awful soldering job was the only thing the previous owner messed with on that machine. I find it so interesting that the II can be upgraded so heavily at a logic board level - were these changes you've made ever manifested in another version of the II?

 

[joshc]

P.S. No, I did not actually throw away the SIMMdoubler right-angle connector.

Phew! You had me for a minute...

Really nice write-up again.

 

[David Cook]

I'm fairly certain they only changed one battery out because the other is for keeping time...and apparently they didn't care outside of the machine starting up.

Interestingly, both batteries are needed for power on.



But only B2 is needed for the clock. They replaced B2. I guess a weak B1 and an overpowered B2 was good enough.

 

[David Cook]

were these changes you've made ever manifested in another version of the II

That's a great question. The Macintosh II continued to ship for over a year after the IIx was released. The IIx had newer ROMs and 1440K drives. I assume late-manufactured Macintosh II computers got the newer (IIx) ROMs from the factory. It is difficult to prove this. I guess if we find a late Macintosh II with new roms but an 800K drive, or we find a late Macintosh II with an old ROM, we could assume that the ROMs were not updated by Apple except through an upgrade kit. Or, maybe we could find the instructions for an upgrade kit that says "if your ROMs have part number xxx, then you only need to swap out the drive".

I would guess that the PMMU never got updated by Apple, just because of cost.

I don't know about the FPU. The Macintosh IIx had a 68882, so there's another opportunity for Apple to update late Macintosh II computers. But I suspect the supply of 68881s were still available at a cheaper price. It would have been a product differentiator.

 

[David Cook]

Really nice write-up again.

Great work and write-up as usual. It's always a pleasure to read your posts.

Thank you. That's very kind of you both.

 

[LaPorta]

Yes this is a very fine write up. I wish I still had the time!

 

[micheledipaola]

Far from pointless! The 68882 is much faster

Important for Excel power users and ensuring your tax returns are done on time.

Maybe this brings us a bit OT, but:

I have a 68881 installed in my Mac II, and a 68882 ready, that I would put once I can get PAL SIMMs and increase RAM over 8mb - BUT :
it is a RC40A, so it should run at 40MHz (instead of the 68881 speed of 16Mhz).

I guess that due to the system / bus speed, it would not run at 40mhz anyway, but I wonder if it could harm to put that onboard instead of a RC16A.

 

[David Cook]

a RC40A, so it should run at 40MHz (instead of the 68881 speed of 16Mhz).

40 MHz is that chip's maximum officially rated speed. As long as the rated speed is equal or more than what the computer will actually run it at, then its good. In your case, the Mac II will run your 40 MHz rated part at 16 MHz. All good.

I guess that due to the system / bus speed, it would not run at 40mhz anyway

Correct.

but I wonder if it could harm to put that onboard instead of a RC16A.

No harm. All good.

 

[micheledipaola]

40 MHz is that chip's maximum officially rated speed. As long as the rated speed is equal or more than what the computer will actually run it at, then its good. In your case, the Mac II will run your 40 MHz rated part at 16 MHz. All good.



Correct.



No harm. All good.

Thanks a lot.

BTW looking for a RC16A 68882, I found RC50A branded chips. I wonder how much legit they are, but in case, it could be that there were faster models in the end.

 

[Melkhior]

it is a RC40A, so it should run at 40MHz (instead of the 68881 speed of 16Mhz).

No, it was certified to run at up to 40 Mhz, but it will run at whatever clock it is fed. In a Mac II, it will be fedt the 16 Mhz clock of the CPU so will run at 16 MHz. Some chips have lower ratings (much less obvious than higher!) below which they can become unreliable, but I think the '881/'882 are all safe to use at 16 MHz.

The '881 and '882 can run asynchronous to the CPU, so from a different clock, up ti a maximum of 2x the CPU according to the documentation. But you would have to modify the mainboard (or use some interposer board) to feed a different clock to the FPU (my IIsi PDS Adapter does that, it's not very difficult).

I guess that due to the system / bus speed, it would not run at 40mhz anyway, but I wonder if it could harm to put that onboard instead of a RC16A.

It's very likely to work, but only at 16 MHz. The '882 is still better than the '881, but while software compatible for userland (regular applications), the OS must support it - it has a different set of state frames than the '881, so early System might choke on it if they don't support the newer state frames (for exception handling, mostly).

Anything supporting the Iix and newer should be fine.

I found RC50A branded chips. I wonder how much legit they are, but in case, it could be that there were faster models in the end.

I believe Motorola did make RC50A '882, but in the same way than Romane-Conti did make a 1945 vintage wine. A heck of a lot more were sold than were actually made! Unless you sourced them from a highly reliable source, or have had them for a long time, they're a significant chance they are fakes (probably '882 but re-labelled).

If they are genuine, I'm interested

 

[David Cook]

it could be that there were faster models in the end.

Yes. In fact, the final DayStar 68030 + 68882 cards ran at 50 MHz and the chips were marked as such. Other brands of accelerators also ran at 50 MHz.

That doesn't mean that someone hasn't remarked 40 MHz chips to look like 50 MHz.... but the 50 MHz versions did exist.

 

[David Cook]

A little bit about binning.

Manufacturers specify the speed range, temperature range, and voltage range of the chips they sell. They then take whatever the worst conditions are for the range (fastest, hottest, lowest?/highest? voltage) and test each individual chip they make as they come off the assembly line.

Let's say they start by testing the 68882 at 52 MHz (a little bit higher than they will mark just to be sure) and the worst temperature/voltage. If it passes, great! They can mark it 50 MHz. Otherwise, try 42 MHz. Otherwise try 36, 27, 18 MHz. They can mark them 40, 33, 25, and 16 respectively.

If almost none of the chips are certified at 50 MHz, it probably isn't worth offering that speed in the catalog. You don't want your customers to be angry that you can't deliver enough 50 MHz chips. So, the manufacturer will have prechecked that they can indeed produce a decent number of the fastest rate chips. They will price these at a premium to maximize profit and keep demand in check. After all, if the price is the same regardless of speed, everyone is just going to order the fastest certified.

Your customers build a bunch of designs and start ordering your chips. Turns out that based on your pricing and the other chips that the customer is relying on in their design that 25 MHz is the most popular speed. But, you've gotten really good at manufacturing and all your chips pass 50 MHz. What do you do? You DON'T mark them all 50 MHz. Instead, you mark enough 50 MHz to meet demand, then mark enough 40 MHz to meet demand, then mark 33, and then mark 25 (the most popular) and finally mark 16. Remember, they all pass the 50 MHz test. But, if you ship a customer a batch of 50 MHz marked chips on a 25 MHz priced order, then why would anyone pay for the 50 MHz chips? I mean, you could occasionally send some faster chips as a 'favor'. But, if you do it enough, the customer will catch on and order the cheapest chips expecting to actually get the best.

Another situation: Let's say Apple orders a million 16 MHz chips. To speed up manufacturing, you decide to skip the tests at faster speeds. I mean, you need to make a million chips for Apple right away! All of these chips might have passed the 50 MHz test, but you didn't even bother trying because you have a big 16 MHz order to fill. Why waste time?

Lastly, let's say the chips are becoming obsolete. You have bunch of 33 MHz chips that actually can run at 50 MHz but you marked them 33 to meet expected demand a couple of years ago. Sonnet comes along and says "I'd like to buy 10,000 50 MHz chips". You don't have that many and you don't want to start up the assembly line for obsolete chips. You can either retest and remark the 33 MHz chips (sometimes they do this) or you can just say "Sonnet -- you test them yourself". You'll notice that the end-of-the-line Sonnet 68040 accelerators often have 33 MHz chips on 40 MHz accelerator boards. That's because the chip was fine at 40 MHz but not marked that way when first produced.

So, the mark only means the chip will go at least this fast. When, in fact, all the chips with the entire range of markings might be exactly the same with all being capable of the fastest speed. This all comes down to pricing and manufacturing efficiency.

 

[micheledipaola]

If they are genuine, I'm interested

AliExpress so I tend to doubt it (and 70€ instead of 20/30€ for the others).