SC Tom wrote:
*** = replies in-line
"VanguardLH" wrote in message
...
SC Tom wrote:
Years and years ago, there used to be these adapters that you could
plug 2 sticks of RAM in to increase the amount memory in each slot
(had some on an old 386 or 486 PC). Are they still being made that
would work on DDR3 RAM, or are the newer memory speeds making that
unstable? I have a number of 8GB sticks around and would love to bump
my desktop from 16GB to 32GB, just for grins 'n giggles (and without
spending big bucks on 2 16GB sticks).
Each mem slot is rated for a maximum capacity based on several factors,
one of which is the number of address lines. Rarely are mobos
overbuilt, and if they were then they would support bigger capacity
modules. You'd have to look at the specs for the mobo where you want to
try this gimmick to see what is its rated max capacity for a module. If
the mobo was designed for 8GB sticks, it doesn't have another address
line to support 16GB sticks.
*** The MB is a BioStar A68-MD Pro with an AMD Athlon(tm) X4 845 Quad
Core Processor (3.5GHz)
From the manual:
"Supports Dual Channel DDR3 800/ 1066/ 1333/ 1600/ 1866/ 2133/
2400(OC)/ 2600(OC)
2 x DDR3 DIMM Memory Slot, Max. Supports up to 32 GB Memory
Each DIMM supports non-ECC 512MB/ 1/ 2/ 4/ 8/ 16 GB DDR3 module"
The RAM I have is all 800MHz (PC3-12800).
You never mentioned the brand and model of your mobo, or which CPU is
installed. Go check their specs to see what is the max size of memory
module is supported. Sticking an adapter card into a slot won't make
another address line magically appear or the controller support larger
sticks. The extra address lines may be physically present, but that
doesn't mean they are logically enabled. Limiting the size of memory is
how makers can lure consumers to spend more on mobos with larger memory
capacity. They can even use the same controller and traces to build
multiple mobo models with differing memory capacities, and just use the
firmware to decide how much gets supported. They economize on the
tooling and parts needed to build the various models, but target
different pricing points for consumers that want to shave a little bit
by not paying for the max capacity.
You also never mentioned which OS you are running on your computer.
Besides limitations (real or throttled) in the hardware, there are
enforced restrictions in max memory capacity that vary by the version of
Windows.
*** I'm running Windows 10 Pro x64.
Windows XP Home: 4 GB
Windows XP Professional: 128 GB
Windows Vista/7 Home Basic: 8 GB
Windows Vista/7 Home Premium: 16 GB
Windows Vista Business/Ultimate/Enterprise: 192 GB
https://docs.microsoft.com/en-us/win...ndows-releases
Shows more Windows versions and their artifical throttle on their
supported maximum memory size.
Like a flooring or least() function, you get the minimum of the maximum
memory capacity between the hardware and the OS.
Thanks for your reply!
https://www.crucial.com/usa/en/compa...star/a68md-pro
Crucial 32GB (2 x 16GB) DDR3L-1600 UDIMM $291.99
CT2K204864BD160B
Some other people, using 16GB sticks on AMD, claimed to have
had problems. A four slot person could get three 16GB sticks
working but not the fourth. Hard to say if that was an
address decode problem or what exactly the problem was there.
You could make a DIMM with a single rank of x4 chips. A company
making DIMMs like that, would use server chips. Whereas Crucial
is less likely to pull a stunt like that. They would hopefully
be x8 chips. But there's no datasheet. Kingston is the
company that provides datasheets with theirs.
And Kingston doesn't list a 16GB DIMM for it. So we can't
use their datasheet and see what is technically on offer.
https://www.kingston.com/us/memory/s...ne=A68MD +Pro
*******
As for combining SPD readouts on the DIMMs, you could do
that with a microcontroller. The PC side SMBUS uses the
low speed option, and a microcontroller could provide register
readout in real time, to cook up a composite SPD value. It's
the details of registered delay (if you use a registered
design), that isn't going to work on a desktop board.
Your hypothetical RAM doubler adapter would be a better
fit in a server case, as it would "tolerate" an exotic
design. I can't see how you would code a UDIMM table,
and have the result work in an RDIMM way. (The Northbridge
CAS delay, has to take into account the one cycle delay
through the register chip. And that's only going to happen
if the BIOS is aware it's an RDIMM. And the keying on the
slots would prevent insertion of an RDIMM as such. In other
words, somebody has to tell the motherboard designer and the
BIOS designer, what the plan is, for the exotic adapter to work.
We couldn't do this on the sly.)
If you didn't use a registered adapter design, then you
would probably overload some of the signals by placing
two DIMMs in one slot.
*******
This crazy idea here works, because it's just an extender
and the wiring is a passive thing. There's got to be
some signal degradation by doing this. Hard to say
when the **** will hit the fan (like, using four of them).
https://linustechtips.com/main/topic...o-yes-you-can/
But the company making that adapter, didn't risk much
by doing it.
Paul