wrote:
Hi all.

Perhaps this is not feasible, but I'd still like to know.
So here goes...

I bought a handy HP 2140 netbook.
Its only drawback is that it's limited to 2GB RAM.

I put a 256 GB SSD and installed ZRAM.
It runs Linux Mint (Tricia) and it's not too bad... as long as I don't have too many things open.

I think it would perform a better if it had 4 GB RAM.
And it has a vacant ExpressCard slot, just sitting there...

So suppose I could find such a gizmo, would it be possible to expand the physical RAM with a chip matrix mounted on an ExpressCard/54 slot or would the bandwidth cause it to bottleneck?

Thanks.
x13

This style of question comes up a lot, but there's really
no movement in industry about the topic.

It's possible the issue is something related to patents,
or the interpretation of what some patent covers, where if
the business opportunity took off, it would be viewed as
patent infringement.

Technically, I think you could make an aid for computer
usage, on plugin cards, which would confir enough advantage
to make it worth installing. But no one is killing themselves
to make it happen.

There was the Gigabyte iRAM (PCI card, interface SATA),
the ACard one (external box, interface SATA) and an absurd
third offering at $10,000 per unit (what were they smoking???).
But none of these had legs. The ACard one stuck around
for several years, whereas the Gigabyte one seems to stop
almost as if someone sent them a letter after the first
batch shipped and told them to stop.

The $10,000 one might have been based on the perception of
"the patent holder wants $9000 licensing per unit", that
sort of thing.

In any case, the thought is there, but no legs.

*******

And the infrastructure was there in the past. We had
Expanded and Extended memory and associated paging mechanism,
so that information could be copied from a specific 4K segment
of a standalone memory, and placed into the main memory
address space. It's not like they didn't build stuff
like this in the past, one way or another.

Every time this question comes up, nobody even has a hint
of an answer as to what's stopping it. Over in sci.electronics,
in the FPGA section, someone might ask whether there'd be
any market if he "built a RAM controller". And nobody seems
to answer questions like that. And the FPGA chips themselves,
are strangely absent the necessities to do banks of RAM
chips. When FPGAs support RAM, they might have a special
facility to support one DRAM chip, but this is useless.
And bodging capabilities out of regular FPGA resources
doesn't usually end up in an "impressively capable" item.
Someone mentioned doing a radar design, where the FPGA
and a RAM buffer held some kind of real time data, but
that's about the closest I know of to "something similar"
in a design.

Sorry to disappoint you, but some "evil force" out there
says "No, no cookie for you".

All that's really required, is a zero seek time paging device,
to extend memory a bit. It doesn't have to be particularly
fast. Even at 4GB per second, it feels "buttery smooth". I've
tested this. But is it good enough to justify any selling
price, or going up against some big company holding
patent(s) ? Probably not. A few people would try it out,
**** on it, and then no one else would buy it.

Intel has done something similar. There are certain SKUs of
Xeon server designs, that you plug in one or more Optane
drives (special flash on a PCI Express card), and those
are used as a memory extension. The idea is, you might
have 256GB of real RAM on a server machine, and the Optane cards
could expand the total virtual space to 2TB. There has to be
a certain ratio of Optane card storage to real RAM, to make
this sorta work. And the software portion (which is what makes it
work), checks to see that the special Xeon is present, and
won't run on older kit or on consumer systems. The Optane
cards themselves aren't that fast, maybe 2200MB/sec, 10us
seek time. And there are probably applications (Exchange Server,
SQL server) that could use it. But again, no interest in
offering it to any except the highest "tier" of device.

Intel has a vested interest in RAM starvation. That's why
they made $17 processors with 1GB RAM limits. The RAM interface
could easily have hosted 4GB or 8GB at least, but they made
sure nobody could build a laptop out of the $17 processor.
And that's called market segmentation. That's why your
machine has its 2GB limitation. It's the processor
just above the $17 one. And so on. A $300 processor uses
whatever RAM fits in the slot, by comparison. That's the
concept with the marketing segmentation. Technically, a $17
processor should be just as capable as a $300 processor,
but they "cut the balls off it", to make sure it's
only "worth" $17. The Intel salesmen don't get those
Gold Rolex watches of theirs, by being nice guys.

Paul