PDA

View Full Version : 486 RAM Bus Wiring...



Raven
February 22nd, 2011, 10:45 AM
The Presario 400 series (Net1,Net1/33,425,433) supports 20MB RAM, normally. It's got 4MB soldered onto the main board, and two sockets that support 8MB sticks each.

I've been theorizing about adding additional RAM sockets - possibly up to eight more for a total of 68MB, to allow the machine to better handle Windows and multitasking, particularly when paired with an Am5x86-133 or P24T - the RAM becomes a bottleneck.

I don't, however, know much about low-level electronics compared to most of you on this forum.

Here's my theory:

It appears that RAM sockets are wired in series - each one has a bunch of traces leading to the same pins on the next, forming a chain of contiguous memory, as though it were one large RAMstick.

If this is the case, you could design expansion units for any given (486) motherboard to add more RAM sockets via plugging a riser card into an existing socket.

I expect there's some reason that this wouldn't work, but I can't see it, so please let me know your thoughts - as I said, you guys are more adept with low-level electronics than I.

glitch
February 22nd, 2011, 11:07 AM
It's not a matter of per-SIMM capacity -- if this were the case, a larger-capacity SIMM would have the same effectiveness as a riser card allowing multiple SIMMS (risers do in fact exist already...I had several that allowed four 30-pin SIMMs to be plugged into a 72-pin socket to keep your previous investment from becoming completely obsolete!). To overcome the RAM limitations of machines like these, you'd have to work around chipset and/or BIOS limitations.

Raven
February 22nd, 2011, 11:50 AM
I've been trying to see if EDO sticks work in this box (without EDO features ofc) but apparently not, either the box won't post, it goes into a partial-post-loop, or it posts to memory errors.

Having given up on that, I was thinking perhaps 16MB SIMMs would work fine in this box, if I can find some FPM ones. It might be that the documenation and testing of this machine was written before 16MB sticks were available. Unfortunately, the vast majority of my 72-pin RAM is unlabeled, even serial numbers or anything I can search for. I could look up the chip sizes, etc., but I've got hundreds of sticks and there are no markings for "1x36" and such things to do calculations with - if I recall properly I've been told that the number of chips correlate in some way, but it doesn't sound fun to look up and calculate capacity for more than a hundred RAM sticks, when I doubt I've got a 16MB FPM stick in the first place...

If there is a hard limitation on stick capacity on this machine, it's the BIOS that imposes the limit, not the chipset. Documentation on the chipset's cache controller makes references to being able to cache up to 256MB of RAM across 1MB L2 cache (missing from the board, I'm looking into that as well). This seems to indicate that the chipset supports at least 256MB of RAM - this is a very compact chipset solution, just two chips (cache controller and "chipset"), a later design, so well possible it supports that much. Unfortunately I haven't located the datasheet for the main chip, as of yet.

A Compaq employee posted this on a forum at some point from some in-house documents:


425 Memory
-----------
System memory of the Compaq Presario 400 Series Personal Computers is based on a 32-bit data path and consists of the following elements:
o 4 MB of 70-ns enhanced page memory standard
o Memory is expandable to 20 MB
o 1, 2, 4, 8 MB industry standard SIMMs may be used

Memory Devices

The Compaq Presario 400 Series Personal Computers come equipped with four MB 70-ns enhanced page mode Dynamic Random Access Memory (DRAM) standard. This memory is soldered to the system board in the form of eight 1-MB x 4 devices. The computer is equipped with four 72-pin SIMM sockets for memory expansion.

NOTE: The first socket must contain a Compaq approved 4-MB SIMM. The remaining three sockets will take a 16-MB SIMM each.


The information here is a bit iffy, because the 400 series only has two sockets. It also is self-contradictory - the first section mirrors what other places say, and the second section describes features that these machines simply don't have, at least not officially or in the final revision.

Basically what that told me is it's well possible that this machine was designed originally to support four sockets. It's also possible that the board supports 16MB SIMMs, or is capable of supporting it given the BIOS doesn't deny that functionality.

RetroHacker_
February 22nd, 2011, 12:11 PM
It's more complicated than just "stuffing more sockets in there". DRAM is addressed differently - you have address and data lines, as well as row and column strobes for each bank. The memory controller needs to be able to communicate with all of the installed RAM - and some of those signals may or may not be present on your board. One thing to do would be to see if you can find the pinouts of the memory controller used on that board.

Another thing to keep in mind, is that 72 pin memory is technically 32 bits wide (36 with parity). 30 pin memory is 8 bits wide (9 with parity). Electrically, the 72 pin modules are addressed very similarly to four individual 30 pin modules - that is, there are four row and column strobes. So, it's really very much like four banks of 8 bit wide memory. IIRC, not all modules were like that - some only contained two banks of RAM, it was the double density ones that had four.

Those socket adapter boards that let you re-use 30 pin memory were pretty popular back then, and came in both front and back configurations - i.e., the 30 pin sockets stuck out either the front or the back, allowing you to use two of 'em in the same machine. There also existed similar boards that allowed you to double up two smaller 72 pin modules in the same slot - but I believe this only worked to allow you to stuff two of the "two bank" modules into one slot as a "four bank" module. I've got a couple of those adapters still, I'll have to see how they are wired.

My memory on this is fuzzy though...

-Ian

Tetrium
February 22nd, 2011, 02:32 PM
Unfortunately, the vast majority of my 72-pin RAM is unlabeled, even serial numbers or anything I can search for.

You can, for a great deal, figure out the memory size by looking at the number of metal legs on each of the memory chips on the SIMM.
For 72p SIMM's, if it has 8/9 chips with 4x 5legs, it's not 16MB. If it has 4x 6legs, it probably is ;)

Agent Orange
February 22nd, 2011, 07:44 PM
It's not a matter of per-SIMM capacity -- if this were the case, a larger-capacity SIMM would have the same effectiveness as a riser card allowing multiple SIMMS (risers do in fact exist already...I had several that allowed four 30-pin SIMMs to be plugged into a 72-pin socket to keep your previous investment from becoming completely obsolete!). To overcome the RAM limitations of machines like these, you'd have to work around chipset and/or BIOS limitations.You could probably figure out a way to phyically mount them on the board (maybe), but I'm not so sure that BIOS will let it happen. Case in point: I have
one 486 mobo w/3 PCI slots and four 72-pin memory slots. The max RAM it supports is 64 MB. I've tried four 32 MB sticks and it won't boot, much less even see them.

Raven
February 23rd, 2011, 05:11 AM
The thing is, I'm not sure if there is a limit or not, as I've only been able to locate >8MB EDO SIMMs so far in my collection, so I haven't been able to check. Perhaps it can only handle 8MB SIMMs but more RAM total - that was the line of thought that brought on this thread. Also, there are replacement BIOSes now, generic BIOSes. Perhaps one of those could be flashed onto a compatible chip and swapped - see if that could be more versatile..? (Unfortunately I don't remember the name this moment, but it was an open source BIOS implementation that could be flashed onto systems)

What about the 4MB of RAM on the system board itself - could chips be piggybacked onto those chips to increase that "SIMM" to 8MB? That would get me to 24MB, a modest improvement. This wouldn't need any added space, physical conversion boards, etc.. Hopefully it wouldn't even need to be soldered, the chips could be "clipped" on until we know if it works, and then removed if not. Thoughts?

I know I'm obsessive and a bit crazy with this stuff, but at least it gets discussion going in interesting directions. xD

Agent Orange
February 23rd, 2011, 07:44 AM
The thing is, I'm not sure if there is a limit or not, as I've only been able to locate >8MB EDO SIMMs so far in my collection, so I haven't been able to check. Perhaps it can only handle 8MB SIMMs but more RAM total - that was the line of thought that brought on this thread. Also, there are replacement BIOSes now, generic BIOSes. Perhaps one of those could be flashed onto a compatible chip and swapped - see if that could be more versatile..? (Unfortunately I don't remember the name this moment, but it was an open source BIOS implementation that could be flashed onto systems)

What about the 4MB of RAM on the system board itself - could chips be piggybacked onto those chips to increase that "SIMM" to 8MB? That would get me to 24MB, a modest improvement. This wouldn't need any added space, physical conversion boards, etc.. Hopefully it wouldn't even need to be soldered, the chips could be "clipped" on until we know if it works, and then removed if not. Thoughts?

I know I'm obsessive and a bit crazy with this stuff, but at least it gets discussion going in interesting directions. xDI don't think that would fly. "Bread boarding' a production mobo is never a good thing unless you're doing it for the sake of science. My thinking is that if they wanted you to have the ability to upgrade, they wouldn't have soldered them to the board in the first place.

Raven
February 23rd, 2011, 08:34 AM
Well if they wanted me to have a P24T they would have given me a socket 2, but I didn't let that stop me and now look what I've got. :D