• Please review our updated Terms and Rules here

what modern RAM upgrade cards are there for 16bit AT class machines currently?

VERAULT

VERAULT

Veteran Member
Joined
Jan 30, 2012
Messages
8,541
Location
Connecticut, USA
I want to soon start my 5170 hercules graphics system restoration... However one point has been nagging at me. What current RAM upgrade cards are there for 16bit AT systems (or newer). There were a bunch of 286 and 386 systems that did not have sim sockets for ram upgrade.

I have a BocaRAM AT PLUS card complete in box however its a 0K model and you need to supply your own RAM chips (72 chips in total) And for a total of up to 8MB I think the cost of the chips would be really prohibitive.

So what are my options currently?
 
Last edited:
There were lots of vendors for 16-bit ISA extended/expanded (configurable for either) memory cards. Everex comes to mind right off the top of my head, say the RAM 10,000 card.

But if you're asking for "modern/current", I don't think there's a large market for those... When the 80386 systems came along, it was all over for 16-bit expanded ISA cards, by and large.
 
Some time ago I got kind of excited over the idea of building a 16 bit card using SRAM or PSRAM, because the challenge of tackling the whole 8/16 bit bus steering and 0WS logic looked interesting, but between being too busy and not actually owning a 286 system to prototype with the enthusiasm waned. (Was planning to use GALs to decode everything, tested some prototype equations, and I have a partially complete schematic in Kicad, but that's as far as it went. If someone wanted it I might be able to publish what I have, but, well, I have no idea if it's any good.)

Per @Chuck(G)'s comment, it is indeed a pretty niche market to build for. Another issue is that while simple old-fashioned brain-dead EMS 3.2 memory is easy to pull off with minimal hardware (and a driver with source for that exists), the more sophisticated EMS 4.0 mappers you'd need to meaningfully run things like Desqview aren't a thing I've seen anyone re-implement yet in the modern day.

If you make a brutally honest assesment of the options that exist today it's kind of hard to not conclude that maybe the best option for maxing out a 286 machine like 5162 or 5170 is to remove the original motherboard, place it in a nice safe static-free padded storage box, and slap in one of those tiny 386SX (or full DX) motherboards that last I checked were still available pretty cheaply on eBay. With the turbo switch disengaged it'll run at almost exactly the same speed as a typical 286, and you'll get SIMM sockets and full memory management along with the deal.
 
Last edited:
Eudimorphodon said:
Some time ago I got kind of excited over the idea of building a 16 bit card using SRAM or PSRAM, because the challenge of tackling the whole 8/16 bit bus steering and 0WS logic looked interesting, but between being too busy and not actually owning a 286 system to prototype with the enthusiasm waned. (Was planning to use GALs to decode everything, tested some prototype equations, and I have a partially complete schematic in Kicad, but that's as far as it went. If someone wanted it I might be able to publish what I have, but, well, I have no idea if it's any good.)

Per @Chuck(G)'s comment, it is indeed a pretty niche market to build for. Another issue is that while simple old-fashioned brain-dead EMS 3.2 memory is easy to pull off with minimal hardware (and a driver with source for that exists), the more sophisticated EMS 4.0 mappers you'd need to meaningfully run things like Desqview aren't a thing I've seen anyone re-implement yet in the modern day.

If you make a brutally honest assesment of the options that exist today it's kind of hard to not conclude that maybe the best option for maxing out a 286 machine like 5162 or 5170 is to remove the original motherboard, place it in a nice safe static-free padded storage box, and slap in one of those tiny 386SX (or full DX) motherboards that last I checked were still available pretty cheaply on eBay. With the turbo switch disengaged it'll run at almost exactly the same speed as a typical 286, and you'll get SIMM sockets and full memory management along with the deal.
Click to expand...
Sounds to me like your mostly there. Where are you located? If it wasnt terribly far id give you a 5170 if it helped produce a card that would benefit the community.
 
I completely forgot about that. Wasn’t I supposed to send you an NEC powermate?
 
I guess at some point if I have the time I should at least shove what I have so far onto GitHub.

FWIW, I dug around Mouser and Digikey, and it looks like for surface-mount (but still reasonably friendly packaged) SRAM populating a board would be in the $3-$4 a megabyte ballpark. There are BGA PSRAMs like this that are only $4 for eight megabytes, but you'd have to have a fab assemble the board. I've vaguely heard that the price for getting that done isn't actually that bad, but I've not done any shopping to try to figure out what that would be for a prototype.

If he can get it to work @ab0tj's design using SIMMs is likely to be a lot cheaper, unless the price of 30 pin SIMMs has gone absolutely bonkers when I wasn't looking.
 
Eudimorphodon said:
I guess at some point if I have the time I should at least shove what I have so far onto GitHub.

FWIW, I dug around Mouser and Digikey, and it looks like for surface-mount (but still reasonably friendly packaged) SRAM populating a board would be in the $3-$4 a megabyte ballpark. There are BGA PSRAMs like this that are only $4 for eight megabytes, but you'd have to have a fab assemble the board. I've vaguely heard that the price for getting that done isn't actually that bad, but I've not done any shopping to try to figure out what that would be for a prototype.

If he can get it to work @ab0tj's design using SIMMs is likely to be a lot cheaper, unless the price of 30 pin SIMMs has gone absolutely bonkers when I wasn't looking.
Click to expand...
With github you can make new 30pin simms(gerbers).
 
VERAULT said:
With github you can make new 30pin simms(gerbers).
Click to expand...

Sure, but you'll need to track down long discontinued Fast Page Mode DRAMs to populate them.

(Which seem to exist in adequate numbers for a few different eBay sellers to have gone into business churning out replicas, which just proves my point that a SIMM-using card is probably the way to go unless you *really* thought there was enough of a market out there for a replica to make using modern BGA memory worth the startup cost. I'm sure if you could make a few hundred of them you could get the price *way* down, but quantity one... no. )
 
You must log in or register to reply here.
Back
Top