64MB EDO RAM in a Color Classic II/ LC550

[beachycove]

In my recent conquest to end all conquests (which it probably will, if my wife has anything to do with it), I acquired 5 x 64MB 72-pin EDO SIMMs. Naturally, I figured that these could be put to interesting uses in some of my old gear.

Thus on Sunday past I popped one of them in my Color Classic + LC550 logic board = Color Classic II. Just for a thrill, you know, since this machine is a particular favourite. Unfortunately, though the machine booted fine, the 64MB SIMM showed up as a 16MB chip (the machine registering a mere 20MB RAM total, with 4MB being on the l.b.). So "that put ma gas on a low peep" (Scots).

Can anybody explain what happened to the other 48MB? Why only 1/4 of the RAM and not, say, 1/2 (which I have heard of before)?

 

[Anonymous Freak]

1. Try a different SIMM. The one you tried may be bad.

2. The LC 550 only supports 32 MB SIMMs. There could be something in the way the 64 MB SIMM was constructed that makes it not 'fail gracefully' to 32 MB, but only all the way down to 16 MB.

 

[johnklos]

Your system can only use 32 meg SIMMs as two banks of 16 meg. Since 64 meg SIMMs are a single bank, you're just seeing memory in one of the two banks, hence 16 megs.

 

[trag]

Your system can only use 32 meg SIMMs as two banks of 16 meg. Since 64 meg SIMMs are a single bank, you're just seeing memory in one of the two banks, hence 16 megs.

What John said.

A 32 MB 72 pin (32 bit wide) SIMM is two banks of 16 MB each. In order to address 16 MB which is 4 bytes wide (32 bits) the Mac must address 16M/4 addresses = 4M addresses. This requires 22 bits of address. The address is multiplexed into two parts, a row and a column. So your Mac supports 11 X 11 addressing. An 11 bit row address is sent to the memory chips, then an 11 bit column address and thence, some address amongst the 4M is found and 4 bytes are obtained from (or written to) that address.

And 11 X 11 is the most that that Mac model knows how to address.

A 64 MB SIMM is one bank of 12 X 12, or 24 bits of address which gives it 16M addresses with 4 bytes per address. But your Mac only knows how to address 22 of those bits, causing a "loss" of 2 bits or a divisor of 4.

A 128 MB SIMM would be seen as 32 MB, because the 128 MB SIMMs are 2 banks of 64MB, just like two 64MB SIMMs pasted together.

 

[johnklos]

Trag - that's as simple an explanation about SIMM sizes as I can imagine. It'd be the kind of thing that should be a sticky or something so that others can find it easily. Instead of reexplaining later, one could simply say, "Go read the SIMM explanation".

How does one request a sticky for a specific post?

 

[beachycove]

Many thanks, and though I don't understand all of what was said (which is a common occurrence for me), I get the general idea well enough. So, before I chuck the RAM in the eBay bin: Are there Quadra-era Macs or early PPCs that 72-pin 64MB EDO chips would work in?

I'd rather not go tearing down a Quadra 800, for instance, if the wee beasties just aren't going to be recognized as 64MB chips in it. On the other hand, if they would work in a Q800, A/UX 3.1 running on a machine with 256MB RAM would make the tear-down well worth the trouble.

 

[johnklos]

Many thanks, and though I don't understand all of what was said (which is a common occurrence for me), I get the general idea well enough. So, before I chuck the RAM in the eBay bin: Are there Quadra-era Macs or early PPCs that 72-pin 64MB EDO chips would work in?

The Quadra 605 (LC 475) motherboard will take 64 and 128 meg SIMMs, as will the PowerMac 6100. From what I've heard, yet without enough confirmation, so will the Quadra 840av.

I'd rather not go tearing down a Quadra 800, for instance, if the wee beasties just aren't going to be recognized as 64MB chips in it. On the other hand, if they would work in a Q800, A/UX 3.1 running on a machine with 256MB RAM would make the tear-down well worth the trouble.

The only way to get A/UX with 256 megs that I know about is the Quadra 950...

 

[trag]

Are there Quadra-era Macs or early PPCs that 72-pin 64MB EDO chips would work in?

Some time ago, I sat down with a pile of RAM (32 MB, 64 MB and 128 MB SIMMs) and a pile of logic boards, Q605, Q610, Q650/800, Q840, PM6100, PM7100 and PM8100. I would have tested a Q630, I have a logic board, but I have no chassis/power supply for the Q630.

I published my results either at 68kmla.org or at Applefritter. However, that may have been before one of the system crashes that lost content at one or both sites...

Fortunately, I keep a lab notebook...

All of the SIMMs worked in the Q605. That's one versatile little machine.

The 32 MB SIMMs worked in all of the machines tested.

The 64 MB SIMMs:

Seen as 32 MB in the Q610 and Q650/800.

Seen as 16 MB in the Q840AV, PM7100 and PM8100.

Works in the PM6100.

The 128 MB SIMMs:

Seen as 64 MB in the Q610 and Q650/800.

Seen as 32 MB in the Q840AV, PM7100, PM8100.

Error Chime in the PM6100.

Reports are that the 128 MB SIMMs should work in the 6100 but mine didn't. However, the way that machines distinguish between banks on a SIMM is by wiring some of the RAS lines to one bank and other RAS lines to the other bank. Then the machine only activates the relevant RAS lines when it wishes to access a particular bank. In the PM 6100 there are four RAS lines which are tied together on the logic board in pairs. On these 128 MB SIMMs the four RAS lines are also tied together in pairs. But the pairs do not correspond. So I suspect that the problem is simply that the RAS lines on these SIMMs aren't wired in a way that is compatible with the way the PM6100 expects.

John, I'm not sure how to get a sticky made. I think you PM the site owner or a moderator -- do we have moderators?

 

[beachycove]

This is encouraging, though the Q800 and 840av tests are a shame. It occurs to me that I have a Q650 in which to test the RAM, just to make sure, and as it is essentially the same as a Q800, I need not go to too much bother to make certain of the A/UX situation.

I do have a Q605 and several LC475s, along with a closely-related machine, the LC575. When I get a few minutes, I will run some tests in these machines (and the Q650) and report back to fill this out a bit. Not any time too soon, mind you. Busy time and the machines are in some cases packed away.