DDR-1 2 gig

[philo]

In my spare parts bin I seem to have a lot of 2 gig sticks of DDR-1

I'm a computer re-builder and must have at least 20 machines in my
workshop that use DDR-1 but none of them support more than 1 gig per
slot. The RAM works fine in the machines but of course only 1 gig (per
slot) is recognized.

Were there any boards that actually support 2 gigs / slot?

If so, I would be worth my while to see if any are listed on eBay.

 

[Paul]

In my spare parts bin I seem to have a lot of 2 gig sticks of DDR-1

I'm a computer re-builder and must have at least 20 machines in my
workshop that use DDR-1 but none of them support more than 1 gig per
slot. The RAM works fine in the machines but of course only 1 gig (per
slot) is recognized.

Were there any boards that actually support 2 gigs / slot?

If so, I would be worth my while to see if any are listed on eBay.

Maybe those sticks came out of a server ?

I'd try to trace down how the modules were constructed. Using
CPU-Z, you can dump the SPD table on the DIMM, then decode
the info in there, to understand it. I have a couple of the
free JEDEC specs that I've used in the past, to do the decoding
step. (What I'm missing now, is the one for DDR3.)

If the modules turn out to cause additional bus loading on
address/control, then they might be poorly suited for desktop
systems anyway. The reason for the additional load, can happen
if a memory chip uses two silicon die inside. There were some
"stacked" modules, and they could have abnormal specs compared
to the regular 1GB modules using x8 wide chips.

For a motherboard to work with the modules, the row and column
address format must be included in the memory controller. It's
possible the 2GB module, requires an additional address bit to be
defined in the controller. Which is why a lot of DDR400 systems
would be stuck at 1GB. If the standards did not define a 2GB
capacity point, that might be why a lot of systems don't support
that large a module. In the past, I didn't notice a lot of
forward-looking thinking, when it came to defining such things.

It still seems to be an issue today, when we have Intel releasing
new processors with twice the memory addressing capability of the
previous DDR3 generation - you'd think JEDEC would include enough
info to handle a few years worth of hardware developments, and
all of that stuff could be included in the first gen of hardware
(i.e. Intel releases 64GB capable processor, when modules
are still limited to a 16GB config, leaving room for future
expansion).

Paul

 

[philo]

Maybe those sticks came out of a server ?

I'd try to trace down how the modules were constructed. Using
CPU-Z, you can dump the SPD table on the DIMM, then decode
the info in there, to understand it. I have a couple of the
free JEDEC specs that I've used in the past, to do the decoding
step. (What I'm missing now, is the one for DDR3.)

If the modules turn out to cause additional bus loading on
address/control, then they might be poorly suited for desktop
systems anyway. The reason for the additional load, can happen
if a memory chip uses two silicon die inside. There were some
"stacked" modules, and they could have abnormal specs compared
to the regular 1GB modules using x8 wide chips.

For a motherboard to work with the modules, the row and column
address format must be included in the memory controller. It's
possible the 2GB module, requires an additional address bit to be
defined in the controller. Which is why a lot of DDR400 systems
would be stuck at 1GB. If the standards did not define a 2GB
capacity point, that might be why a lot of systems don't support
that large a module. In the past, I didn't notice a lot of
forward-looking thinking, when it came to defining such things.

It still seems to be an issue today, when we have Intel releasing
new processors with twice the memory addressing capability of the
previous DDR3 generation - you'd think JEDEC would include enough
info to handle a few years worth of hardware developments, and
all of that stuff could be included in the first gen of hardware
(i.e. Intel releases 64GB capable processor, when modules
are still limited to a 16GB config, leaving room for future
expansion).

Paul

They are definitely for a desktop.
Since I made this post I did find a motherboard that ...according to
it's spec sheet...supports 2 gig sticks for a total of 8 gigs.

I need to build a machine for a friend of mine who is editing *huge*
file-sized images in Photoshop...and he is on a severe budget.

I have all the parts I need already, RAM, CPU etc...
just need the mobo. When it gets here I will see if the RAM is fully
supported. I am using it now in my own machine but it's only recognized
at 1 gig per stick...to to the bios limitation.

 

[Paul]

They are definitely for a desktop.
Since I made this post I did find a motherboard that ...according to
it's spec sheet...supports 2 gig sticks for a total of 8 gigs.

I need to build a machine for a friend of mine who is editing *huge*
file-sized images in Photoshop...and he is on a severe budget.

I have all the parts I need already, RAM, CPU etc...
just need the mobo. When it gets here I will see if the RAM is fully
supported. I am using it now in my own machine but it's only recognized
at 1 gig per stick...to to the bios limitation.

OK, for fun, while we wait for the mobo, post the SPD table.

In CPU-Z, go to the About menu item on the right, and there
is a "Save Report" or "Register Dump" option. In there, look
for the table after "Dump Module #1" etc.

In one version of CPU-Z, the table looks like this.

DIMM # 1
SPD registers
00 01 02 03 04 05 06 07 08 09 0A 0B 0C 0D 0E 0F
00 80 08 08 0E 0A 01 40 00 05 25 40 00 82 08 00 00
10 0C 08 38 01 02 00 03 3D 50 50 60 32 1E 32 2D 01
20 17 25 05 12 3C 1E 1E 00 36 39 7F 80 14 1E 00 00
...
F0 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00

In another version of CPU-Z, the table looks like this.
Same data, slightly different formatting. There should
be 256 bytes of data shown, although most of the valuable
information is in the first part of the table. On good
quality DIMMs, each DIMM has a unique serial number.
On the generics, the SPD chips come out of a bucket
at the plant, all the same (and potentially read-only),
so that you can't tell the DIMMs apart later. In the very
worst of products, the SPD contents are even wrong - one
poster to the groups, presented a table that causes the
BIOS to only use the first half of the DIMM. So there are
a few "village idiots" making DIMMs out there.

Dump Module #1
0 1 2 3 4 5 6 7 8 9 A B C D E F
00 80 08 08 0E 0A 01 40 00 05 25 40 00 82 08 00 00
10 0C 08 38 01 02 00 03 3D 50 50 60 32 1E 32 2D 01
20 17 25 05 12 3C 1E 1E 00 36 39 7F 80 14 1E 00 00
....
F0 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00

The BIOS uses two methods to commission DIMMs. It
trusts the contents of the SPD, but "verifies"
the result, using the ancient probe (peek n' poke)
method. If the SPD is wrong, the BIOS can in a
lot of cases, still figure it out and make the
machine run.

Paul

 

[BW]

In my spare parts bin I seem to have a lot of 2 gig sticks of DDR-1

I'm a computer re-builder and must have at least 20 machines in my
workshop that use DDR-1 but none of them support more than 1 gig per
slot. The RAM works fine in the machines but of course only 1 gig (per
slot) is recognized.

Were there any boards that actually support 2 gigs / slot?

If so, I would be worth my while to see if any are listed on eBay.

--https://www.createspace.com/3707686

I can't recall any Pentium 4 DDR mainboards that handled more that 1
GiB per slot.
I believe that socket 939 Opterons supported unbuffered DDR 2 GiB
modules, maybe that is what they were made for.
I think most servers were built with registered RAM for stability.
A few might have chosen the unbuffered RAM for a cycle less latency,
so there might not be too many buyers in
need of this stuff. But if you have 2 or 4 matching modules, might be
worth a try.

 

[philo]

Seems to have been a foolish error on my part.

I purchased a fairly inexpensive (used) Abit AN8 mobo on eBay

I put in the RAM and saw it was recognized at 1 gig per stick.

When I looked again at the box the two RAM sticks came in it was labeled
as 2gigs 1 gig per stick so the RAM I had was not what I thought
it had been.

Worse still, since it had an older bios it did not correctly identify
the dual core CPU.

None of ABIT's bios flashes were any good...all of them were corrupted
but I went to another site and got a new BIOS flash that worked and
recognized the CPU correctly but the OS crashed several times during my
tests.

I ended up putting in a single core CPU and the thing is now rock stable.

The only good news is that they person I am building this machine for
has almost no money and it will at least hold him until he can save up
for something really decent.


His present machine is 32bit and the maximum RAM that can be assigned to
Photoshop is 1.7 gigs.

One the machine I am building, since I am using the 64 bit version of XP,
I can assign about 3.4 gigs of RAM to Photoshop...
so he should see an improvement.


Anyway, once he gets some money saved, I will be looking for an all new
mobo which will for sure be DDR-3

 

[Paul]

Seems to have been a foolish error on my part.

I purchased a fairly inexpensive (used) Abit AN8 mobo on eBay

I put in the RAM and saw it was recognized at 1 gig per stick.

When I looked again at the box the two RAM sticks came in it was labeled
as 2gigs 1 gig per stick so the RAM I had was not what I thought
it had been.

Worse still, since it had an older bios it did not correctly identify
the dual core CPU.

None of ABIT's bios flashes were any good...all of them were corrupted
but I went to another site and got a new BIOS flash that worked and
recognized the CPU correctly but the OS crashed several times during my
tests.

I ended up putting in a single core CPU and the thing is now rock stable.

The only good news is that they person I am building this machine for
has almost no money and it will at least hold him until he can save up
for something really decent.


His present machine is 32bit and the maximum RAM that can be assigned to
Photoshop is 1.7 gigs.

One the machine I am building, since I am using the 64 bit version of XP,
I can assign about 3.4 gigs of RAM to Photoshop...
so he should see an improvement.


Anyway, once he gets some money saved, I will be looking for an all new
mobo which will for sure be DDR-3

The third customer review posting here, claims to be running an X2
processor on the Abit AN8. The problem would be finding the file
he used, whatever it was.

http://www.newegg.com/product/product.aspx?Item=N82E16813127197

Since I'd be unlikely to get a BIOS here...

ftp://ftp-usa.abit.com.tw/pub/download/bios/an8/

you could always try a hacked BIOS file...

http://www.lejabeach.com/ABIT/an8/abitan8bios.html

*******

On a default 32 bit installation, the virtual memory is split with
2GB address space for kernel addresses, and 2GB address space for
application addresses. This could be the reason Photoshop will
limit itself to 1.7GB. I think I've seen a 1.8GB limit here, and
even with 4GB installed, the same limit is evident.

To make more possible, two things are needed.

1) The /3Gb switch in boot.ini

http://msdn.microsoft.com/en-us/library/windows/hardware/gg487508.aspx

multi(0)disk(0)rdisk(0)partition(2)\WINNT="WinXP descriptive string" /3GB

2) The application itself, must have the "large address aware" flag.
The developer sets that, as far as I know. From the same article...

"Executables that can use the 3-GB address space are required to have
the bit IMAGE_FILE_LARGE_ADDRESS_AWARE set in their image header.
If you are the developer of the executable, you can specify a
linker flag (/LARGEADDRESSAWARE)."

I've never managed to get LARGEADDRESSAWARE working here (i.e. a
working test case), but you're good at this stuff and I'm sure
it'll work for you. I see no reason for Photoshop to not be
compiled that way.

So when you get 4GB of RAM installed, that's not the end of the
story on an x32 OS.

*******

The permutations of LARGEADDRESSAWARE are covered in more detail
here, so you can see what limits might exist with 32 bit Photoshop
on a 64 bit OS.

http://msdn.microsoft.com/en-us/library/aa366778(VS.85).aspx

Paul

 

[philo]

The third customer review posting here, claims to be running an X2
processor on the Abit AN8. The problem would be finding the file
he used, whatever it was.

http://www.newegg.com/product/product.aspx?Item=N82E16813127197

Since I'd be unlikely to get a BIOS here...

ftp://ftp-usa.abit.com.tw/pub/download/bios/an8/

you could always try a hacked BIOS file...

http://www.lejabeach.com/ABIT/an8/abitan8bios.html

As I mentioned *every* bios on Abit's website was defective.
The newest BIOS they had was version AN820.

I did find an apparently 3rd party version AN821 which seemed to have
worked fine in that the dual core CPU was correctly recognized...
but there were several problems running a dual core CPU.
With the single core CPU the system is rock stable so I am going to
leave well enough alone.
My main concern was getting a 64 bit machine, and that's now been
accomplished.

*******

On a default 32 bit installation, the virtual memory is split with
2GB address space for kernel addresses, and 2GB address space for
application addresses. This could be the reason Photoshop will
limit itself to 1.7GB. I think I've seen a 1.8GB limit here, and
even with 4GB installed, the same limit is evident.

To make more possible, two things are needed.

1) The /3Gb switch in boot.ini

http://msdn.microsoft.com/en-us/library/windows/hardware/gg487508.aspx

multi(0)disk(0)rdisk(0)partition(2)\WINNT="WinXP descriptive string"
/3GB

2) The application itself, must have the "large address aware" flag.
The developer sets that, as far as I know. From the same article...

"Executables that can use the 3-GB address space are required to have
the bit IMAGE_FILE_LARGE_ADDRESS_AWARE set in their image header.
If you are the developer of the executable, you can specify a
linker flag (/LARGEADDRESSAWARE)."

I've never managed to get LARGEADDRESSAWARE working here (i.e. a
working test case), but you're good at this stuff and I'm sure
it'll work for you. I see no reason for Photoshop to not be
compiled that way.

So when you get 4GB of RAM installed, that's not the end of the
story on an x32 OS.

*******

The permutations of LARGEADDRESSAWARE are covered in more detail
here, so you can see what limits might exist with 32 bit Photoshop
on a 64 bit OS.

http://msdn.microsoft.com/en-us/library/aa366778(VS.85).aspx

Paul

I have tried on several machines to use the /3GB switch on 32bit XP
but still the system only sees about 2.8 gigs even though in the control
panel it states PAE.

On the machine I am using now, I dual boot 32bit XP and 32bit Linux
and using the PAE kernel in Linux all 4 gigs is recognized.
Do you knwo if there is anything else I can do to get 32bit XP to "see"
all the RAM?

 

[Paul]

As I mentioned *every* bios on Abit's website was defective.
The newest BIOS they had was version AN820.

I did find an apparently 3rd party version AN821 which seemed to have
worked fine in that the dual core CPU was correctly recognized...
but there were several problems running a dual core CPU.
With the single core CPU the system is rock stable so I am going to
leave well enough alone.
My main concern was getting a 64 bit machine, and that's now been
accomplished.

I have tried on several machines to use the /3GB switch on 32bit XP
but still the system only sees about 2.8 gigs even though in the control
panel it states PAE.

On the machine I am using now, I dual boot 32bit XP and 32bit Linux
and using the PAE kernel in Linux all 4 gigs is recognized.
Do you knwo if there is anything else I can do to get 32bit XP to "see"
all the RAM?

The /3GB switch, is so the ~3GB free that WinXP reports, can be used
by one single program. I have a 4GB machine here, and to use 3GB of RAM
requires me to use two or more programs. I have no single program
here, which is LARGEADDRESSAWARE and also has a hunger for RAM.
If I had a LARGEADDRESSAWARE program here, and I set the /3GB switch,
then that program could grow to fill 3GB.

The problem with memory hoisting, is when you install 4GB physical RAM,
the first 3GB are linearly mapped, and the fourth gigabyte is lifted
to between the 4GB and 5GB mark. Something like that. Even the RAMDisk
program, doesn't like that particular segment of RAM, and the RAMDisk will use
RAM either below (AWE) or above (PAE) that mark. I don't know what is magic
about the 4GB to 5GB area. If you watch a modern version of memtest86+
testing RAM, it exhibits the same strange behavior. Segments above 4GB
being tested 2GB at a time, while one segment seems to be 1GB in size.
And perhaps that 1GB segment, is the 4GB to 5Gb area.

PAE is enabled on WinXP, at least for SP3. That's done so they could
fit the No Execute function for malware hardening. So the PAE part
of the solution is there, that would allow more memory in a 32 bit OS.
PAE would not allow a single program to use all the memory either.
There would still be a 4GB virtual address space available to each
process, which bounds how much RAM it could use. If PAE made 64GB
available to me, I'd need more than 16 programs to use it all.
A 32 bit copy of Photoshop, could not use the entire 64GB. Only
a small portion.

The article here, the author of this article, hacks Vista x32, so that
it will use 8GB of RAM. You would need to find the WinXP equivalent of
this, to test out the idea (I'm pretty sure, it's not as easy as the
Vista case). At least we know Win98 is not a candidate,
as Win98 falls apart if offered too much RAM.

http://www.geoffchappell.com/notes/windows/license/memory.htm

OK, here's a hint for WinXP. No idea if this is stable or not.

http://superuser.com/questions/292207/is-there-any-way-to-use-memory-above-3-25gb-using-windows-xp

"I have heard tales that it's possible to break this barrier in
32bit XP, but it requires much more than a simple registry edit.
IIRC, the limit is compiled into the operating system directly.
To get around it, you have to find a specific .dll file from a
32bit Server 2003 machine and use it to replace the equivalent
file on your Windows XP machine. For this to work, the file has
to be modified so that XP won't reject it and you have to use
volume shadow copy to get it to replace the existing file. I don't
remember and can't find the link now for which file you need or
how to modify it. After this is accomplished, you should be able
to make the same settings to 32bit XP that you can to Server 2003
to allow the higher memory cap. Of course, such a change is highly
unsupported and violates your license agreement."

HTH,
Paul

 

[philo]

On 03/02/2013 07:06 PM, Paul wrote:

The /3GB switch, is so the ~3GB free that WinXP reports, can be used
by one single program. I have a 4GB machine here, and to use 3GB of RAM
requires me to use two or more programs. I have no single program
here, which is LARGEADDRESSAWARE and also has a hunger for RAM.
If I had a LARGEADDRESSAWARE program here, and I set the /3GB switch,
then that program could grow to fill 3GB.

The problem with memory hoisting, is when you install 4GB physical RAM,
the first 3GB are linearly mapped, and the fourth gigabyte is lifted
to between the 4GB and 5GB mark. Something like that. Even the RAMDisk
program, doesn't like that particular segment of RAM, and the RAMDisk
will use
RAM either below (AWE) or above (PAE) that mark. I don't know what is magic
about the 4GB to 5GB area. If you watch a modern version of memtest86+
testing RAM, it exhibits the same strange behavior. Segments above 4GB
being tested 2GB at a time, while one segment seems to be 1GB in size.
And perhaps that 1GB segment, is the 4GB to 5Gb area.

PAE is enabled on WinXP, at least for SP3. That's done so they could
fit the No Execute function for malware hardening. So the PAE part
of the solution is there, that would allow more memory in a 32 bit OS.
PAE would not allow a single program to use all the memory either.
There would still be a 4GB virtual address space available to each
process, which bounds how much RAM it could use. If PAE made 64GB
available to me, I'd need more than 16 programs to use it all.
A 32 bit copy of Photoshop, could not use the entire 64GB. Only
a small portion.

The article here, the author of this article, hacks Vista x32, so that
it will use 8GB of RAM. You would need to find the WinXP equivalent of
this, to test out the idea (I'm pretty sure, it's not as easy as the
Vista case). At least we know Win98 is not a candidate,
as Win98 falls apart if offered too much RAM.

http://www.geoffchappell.com/notes/windows/license/memory.htm

OK, here's a hint for WinXP. No idea if this is stable or not.

http://superuser.com/questions/292207/is-there-any-way-to-use-memory-above-3-25gb-using-windows-xp


"I have heard tales that it's possible to break this barrier in
32bit XP, but it requires much more than a simple registry edit.
IIRC, the limit is compiled into the operating system directly.
To get around it, you have to find a specific .dll file from a
32bit Server 2003 machine and use it to replace the equivalent
file on your Windows XP machine. For this to work, the file has
to be modified so that XP won't reject it and you have to use
volume shadow copy to get it to replace the existing file. I don't
remember and can't find the link now for which file you need or
how to modify it. After this is accomplished, you should be able
to make the same settings to 32bit XP that you can to Server 2003
to allow the higher memory cap. Of course, such a change is highly
unsupported and violates your license agreement."

HTH,
Paul

Moving forward I am going to be using 64bit operating systems from now
on and will not have to worry about PAE...especially since there are
finally 64bit applications now available