Slightly OT: A challenge...

  • Thread starter Thread starter Cerridwen
  • Start date Start date
C

Cerridwen

All you need are four things: -

A 386-based system
A copy of Windows XP (flavour doesn't matter)
Time
Patience

Why am I issuing this challenge? Well, there's a bloke, posting in one of
the MS-hosted XP groups, who's insisting he has XP running (and running
well) on a 386. Not to mention the raw clock speed issue, (XP requires a
minimum of 233), but what was the maximum amount of RAM a 386 mobo could
take (whatever it was, I feel certain it wasn't as much as 64MB, which is
the minimum XP requires)?

I'm afraid there are no prizes - just kudos and admiration.
 
Cerridwen said:
All you need are four things: -

A 386-based system
A copy of Windows XP (flavour doesn't matter)
Time
Patience

Why am I issuing this challenge? Well, there's a bloke, posting in one of
the MS-hosted XP groups, who's insisting he has XP running (and running
well) on a 386. Not to mention the raw clock speed issue, (XP requires a
minimum of 233), but what was the maximum amount of RAM a 386 mobo could
take (whatever it was, I feel certain it wasn't as much as 64MB, which is
the minimum XP requires)?

I'm afraid there are no prizes - just kudos and admiration.
The 386 itself could address up to 4GB of RAM, however, mobos of those days
used 30 pin SIMMs and I don't remember ever seeing one with more than 4MB on
it. I have a 25MHz 386DX mobo here with 16 SIMM sockets but only 3x 4MB
memory sticks so its a no-go. In any case I don't think any 386 mobos could
possibly control HDDs of more than 528MB (no LBA in those days) and that's
not enough for even a very basic WinXP installation.
 
Alien said:
The 386 itself could address up to 4GB of RAM, however, mobos of
those days used 30 pin SIMMs and I don't remember ever seeing one
with more than 4MB on it. I have a 25MHz 386DX mobo here with 16 SIMM
sockets but only 3x 4MB memory sticks so its a no-go. In any case I
don't think any 386 mobos could possibly control HDDs of more than
528MB (no LBA in those days) and that's not enough for even a very
basic WinXP installation.

Precisely - here's the thread if you're remotely interested
http://tinyurl.com/2546p. Prior to XP, he claims to have had 2000 installed!
Even if the '3' should have read '4' it would still be an impossibility!

I'd love to see the system - must be some fantastic customisations!
 
Cerridwen's log on stardate 16 ožu 2004
Why am I issuing this challenge? Well, there's a bloke, posting in
one of the MS-hosted XP groups, who's insisting he has XP running
(and running well) on a 386.

Running, yes. Running well, he is a dam liar.
Not to mention the raw clock speed issue, (XP requires a minimum of
233),but what was the maximum amount of RAM a 386 mobo could take
(whatever it was, I feel certain it wasn't as much as 64MB, which is
the minimum XP requires)?

Amptron DX-5500, 386DX 40 MHz, 32 MB RAM, 256KB cache and NPU
(cooprocessor). Drive is not an issue (drive managament utilityes), but
since it requiers time (and i don't have plenty now), it'll have to wait.
However, i hope to do it soon, and i'll post the results.
 
Alien said:
The 386 itself could address up to 4GB of RAM, however, mobos of
those days used 30 pin SIMMs and I don't remember ever seeing one
with more than 4MB on it.

I had a pair of 30 pin SIMMs that were 64MB each, from an early model server
(I bought it for the RAM, 25Mhz 486/EISA based, $20 and it still worked
perfectly), so they were about. Must have cost thousands of dollars when
they were new though. (I remember when 72 pin EDO RAM cost around $NZ100 per
MB) I also had a pair of 72 pin EDO that were 128MB each. (Also from an old
server, I bought that one as it had a relatively new Adaptec PCI SCSI card
in it I wanted) Sold 'em on line auction, the guy who bought the 30 pin used
'em (or one of 'em) in an old SCSI RAID card, the 72 pin went to a guy who
wanted one of them for an Amiga or similar, he said the specs said it maxed
out at 128MB but the most he'd found up 'til then was 8MB. I got reasonable
money for them but not even a tenth (or even a twentieth) of what they would
have cost new.
 
The 386 itself

? You mean the chipset...correct? And what chipset do you remember
that could address up to 4 gig of RAM?
could address up to 4GB of RAM,

I must have been sleeping...or smoking...during that era! lol I
don't remember any of that.
however, mobos of those days
used 30 pin SIMMs

And in 4-stick banks.
and I don't remember ever seeing one with more than 4MB on
it.

I think I remember seeing 4mb sticks toward the end of that
era...maybe even 8's. That would kick the limit to 32...but that's
all I remember.
I have a 25MHz 386DX mobo here with 16 SIMM sockets but only 3x 4MB
memory sticks so its a no-go. In any case I don't think any 386 mobos could
possibly control HDDs of more than 528MB (no LBA in those days) and that's
not enough for even a very basic WinXP installation.

Boy, its hard to remember that far back. But I think my first 386
didn't even HAVE a controller built into the mainboard. But I kinda
remember that it was a 1.2 or 1.6 gig.

I think I'll do some browsin'...just for old-times sake.


Have a nice week...

Trent

Follow Joan Rivers' example --- get pre-embalmed!
 
~misfit~ said:
I had a pair of 30 pin SIMMs that were 64MB each, from an early model server
I've Googled for it and the largest I found was 16MB. There was some talk of
32MB but no evidence of its existence.
 
Trent© said:
? You mean the chipset...correct? And what chipset do you remember
that could address up to 4 gig of RAM?


I must have been sleeping...or smoking...during that era! lol I
don't remember any of that.
No Trent, chipset does not generate memory addresses, the CPU does. Intel
386, 486 and Pentium CPUs have 32 bit address registers. 2^32 is 4GB. The
chipset (the memory controller part of it) buffers the address lines and
generates the timing signals, refresh cycles and parity or error checking.
Its the driving ability of the controller that limits the number of memory
sockets or banks a motherboard can have.
 
No Trent, chipset does not generate memory addresses, the CPU does.

I'm not talkin' about memory addresses, per se. I'm talkin' about the
ability to do what yer sayin'.

I don't think those chipsets (or BIOS in total, if that's the way you
want to look at it) was able to do what you say...even if the hardware
had even been available at the time.

Anyway, those days are long gone! lol


Have a nice week...

Trent

Follow Joan Rivers' example --- get pre-embalmed!
 
No Trent, chipset does not generate memory addresses, the CPU does. Intel
386, 486 and Pentium CPUs have 32 bit address registers. 2^32 is 4GB. The
chipset (the memory controller part of it) buffers the address lines and
generates the timing signals, refresh cycles and parity or error checking.
Its the driving ability of the controller that limits the number of memory
sockets or banks a motherboard can have.

As a good example for my other post...

I'm running a Pentium-class machine with a TX chipset...with DIMM's.
The maximum RAM I can use is nowhere NEAR 4 gig.

Its determined by the chipset.

And I really don't remember what the max was for the 386-class
machines.


Have a nice week...

Trent

Follow Joan Rivers' example --- get pre-embalmed!
 
Alien said:
I've Googled for it and the largest I found was 16MB. There was some
talk of 32MB but no evidence of its existence.

Well, I'm not lying to you. They were 64MB each. Big, tall modules with
double rows of vertical chips on each side and a third row on it's side.
Huge things. I tested them in a 486DX100 motherboard and it showed 128MB. I
also Googled them when I had them, the numbers on the modules, the numbers
on the chips (I found them) but found no evidence of the modules themselves.
Google is good but isn't God.

Actually they were more trouble than they were worth. At the time I had a
bit of cash and was buying up old stuff at auction and breaking it down and
selling components on-line. I thought these would be sought-after by someone
with maybe an old dedicated machine who needed to upgrade. I put the reserve
at $NZ30 and twice they 'sold' only to have the buyer pull out. (Bloody kids
on auction sites) After listing them probably eight times with a $30 reserve
a guy won the auction at $30 and I sent him the modules. However he emailed
me back saying they wouldn't work in his board and he wanted his money back.
He said he'd send them back if I paid postage. By then I was sick of them
and told him to keep them anyway, I just refunded his money. (I have a 100%
satisfation rating on the auction site and I didn't want to lose it, I'd
already had to go through the hassle of recouping the commision fees for the
damn things twice).

I wish I'd kept 'em now, I collect old CPUs, if these were as rare as they
now seem to be they could have been worth hanging on to.

Incidently, I got $NZ210 for the pair of 72 pin 128MB modules.
 
Well, I'm not lying to you. They were 64MB each. Big, tall modules with
double rows of vertical chips on each side and a third row on it's side.
Huge things. I tested them in a 486DX100 motherboard and it showed 128MB.

30 pin required 4 modules in order to work. How did you get them to
work?


Have a nice week...

Trent

Follow Joan Rivers' example --- get pre-embalmed!
 
Trent© said:
30 pin required 4 modules in order to work. How did you get them to
work?

I put them in the first two slots and Viola! It booted and said 128MB RAM.
If it required four modules how come there were six slots? In fact, I've
used a few 30 pin RAM boards that I've used pairs of modules in. Would you
like to point me to a web site that explains that what I did was impossible
please? I would very much appreciate it as I *know* I'm fu*king good but I
didn't know I could work miracles.
 
~misfit~ said:
I put them in the first two slots and Viola! It booted and said 128MB
RAM. If it required four modules how come there were six slots? In
fact, I've used a few 30 pin RAM boards that I've used pairs of
modules in. Would you like to point me to a web site that explains
that what I did was impossible please? I would very much appreciate
it as I *know* I'm fu*king good but I didn't know I could work
miracles.

Just did a web search on this and it seems that most 30 pin RAM did have to
be installed in fours. However, these were double-sided modules and late
model mobos.
 
~misfit~ said:
Just did a web search on this and it seems that most 30 pin RAM did have to
be installed in fours. However, these were double-sided modules and late
model mobos.
SIMMs are so called because the edge connector contacts on both sides of the
stick are electrically joined via a plated through hole. Also the socket
contact is a single piece of metal touching both sides of the stick. If they
weren't joined they would be called DIMMs. So a double sided SIMM still only
has the same width data bus as a single sided.

386SX boards required 30 pin sticks in pairs, 386DX and 486 in quads. There
was a Pentium chipset which allowed a board to work from a single 72 pin
SIMM (Pentiums have 64 bit memory data bus so normally a pair of 72 pin
SIMMs is required). It did this by accessing two consecutive memory
locations for every RAM access, however, this slowed down the PC quite a
bit. I'm not aware of any similar chipset for 386 or 486 boards that would
allow the boards to work with half the normal complement of RAM.

Anyway, all this talks and challenges have made me look through all the long
forgotten boxes only to find one with some 20 different 386, 486 and Pentium
boards. One had four 4MB 30 pin SIMMs and one 32MB 72 pin SIMMs so I can
have a bit of fun this weekend trying to get XP to run on yesteryear's
hardware.
 
Trent© said:
30 pin required 4 modules in order to work. How did you get them to work?

Not always the Case. I have a 486 SX 40 with only two 4MB sticks and it runs
just fine. Wish I could find some more. Its a six slot board with a 6 1/2 X 8
3/4 footprint Says M396F V2.6 Made in Taiwan. I use it as a breadbord for
testing.
Bob in Wisconsin
 
Robert said:
Not always the Case. I have a 486 SX 40 with only two 4MB sticks and
it runs just fine. Wish I could find some more. Its a six slot board
with a 6 1/2 X 8 3/4 footprint Says M396F V2.6 Made in Taiwan. I use
it as a breadbord for testing.
Bob in Wisconsin

Thanks Bob. I was beginning to think I dreamed the whole thing.
 
Alien said:
SIMMs are so called because the edge connector contacts on both sides
of the stick are electrically joined via a plated through hole. Also
the socket contact is a single piece of metal touching both sides of
the stick. If they weren't joined they would be called DIMMs. So a
double sided SIMM still only has the same width data bus as a single
sided.

<Shrug>. Not saying you're wrong, web sites can be wrong too but that's not
what a RAM site I looked at last night said. It said that most 30 pin RAM
was single-sided but some later sticks had chips on both sides and some
boards couldn't address it.
386SX boards required 30 pin sticks in pairs, 386DX and 486 in quads.
There was a Pentium chipset which allowed a board to work from a
single 72 pin SIMM (Pentiums have 64 bit memory data bus so normally
a pair of 72 pin SIMMs is required). It did this by accessing two
consecutive memory locations for every RAM access, however, this
slowed down the PC quite a bit. I'm not aware of any similar chipset
for 386 or 486 boards that would allow the boards to work with half
the normal complement of RAM.

I'm sure I didn't dream it. It was an Osbourne PC, made in Australia.
Anyway, all this talks and challenges have made me look through all
the long forgotten boxes only to find one with some 20 different 386,
486 and Pentium boards. One had four 4MB 30 pin SIMMs and one 32MB 72
pin SIMMs so I can have a bit of fun this weekend trying to get XP to
run on yesteryear's hardware.

Good luck.
 
Back
Top