DDR-1 ?

  • Thread starter Thread starter philo 
  • Start date Start date
P

philo 

I am upgrading some older P-4 and AMD equivalent machines and recently
purchased some dirt cheap Chinese RAM on eBay. ( 1gig sticks 400mhz)


The site specifically said it works with VIA, SiS and Nvidia3/4 chipsets

and indeed it does...so all is fine


however I got documentation with the RAM stating that it does not work
with Intel, Mac or MSI motherboards etc.


Just for the heck of it I tried some in an Intel chipset machine and it
did not even post or give an error message.


Just curious what kind of odd ball ram this is. Up until now any DDR-1
400mhz I've had worked in any machine I've tried it in.
 
philo said:
I am upgrading some older P-4 and AMD equivalent machines and recently
purchased some dirt cheap Chinese RAM on eBay. ( 1gig sticks 400mhz)


The site specifically said it works with VIA, SiS and Nvidia3/4 chipsets

and indeed it does...so all is fine


however I got documentation with the RAM stating that it does not work
with Intel, Mac or MSI motherboards etc.


Just for the heck of it I tried some in an Intel chipset machine and it
did not even post or give an error message.


Just curious what kind of odd ball ram this is. Up until now any DDR-1
400mhz I've had worked in any machine I've tried it in.

They're a 16 chip single rank setup using x4 chips.

The Intel chipset documentation, documents accepted x8 chip and x16
chip configurations. On those, they're arranged in two ranks on
the UDIMM. As an example of a typical DIMM (all my DDR400 are
this way), they're two ranks, 8 chips having x8 width on
each rank, and each rank occupies one side of the DIMM.

The DIMM data bus is 64 bits wide. A rank seeks to match that
width. A DIMM can have multiple ranks, only one of which
is chip_selected during access.

On registered server DIMMs (RDIMM), you can have as many
as four ranks. The register chips provide buffering and fanout
for driving larger arrays, at the price of an extra clock
cycle of latency in the address/control path. Typically a
registered DIMM will use x4 chips, as a means of achieving
a higher capacity DIMM. So that's where the actual market
for x4 chips is, rather than the not-so-compatible 1GB
DDR400 high density DIMM.

Inside each memory chip, they do some similar things. They
have "banks" in there, or rectangles of memory bits. And
the third dimension inside the chip is a "bank". That's why
"rank" is used when describing the array at the memory
chip level on the DIMM. Since the name "bank" was
already taken :-)

*******

What's funny about the DIMM you just bought (16 chip x4 width),
is no company wants to document one of those :-) If I buy
a DIMM from Kingston, I can get a datasheet, where I
can see I'm getting two ranks, of 8 chips of 8x width.

But the generic companies who make the unbranded UDIMM
with the x4 chips on it, they *never* seek to document what
they've done. Instead, they couch their product
description in terms of "works with VIA, SIS, NVidia, and
not so much with Intel". Leaving the customer to
figure out how they've been snookered.

Paul
 
They're a 16 chip single rank setup using x4 chips.

The Intel chipset documentation, documents accepted x8 chip and x16
chip configurations. On those, they're arranged in two ranks on
the UDIMM. As an example of a typical DIMM (all my DDR400 are
this way), they're two ranks, 8 chips having x8 width on
each rank, and each rank occupies one side of the DIMM.

The DIMM data bus is 64 bits wide. A rank seeks to match that
width. A DIMM can have multiple ranks, only one of which
is chip_selected during access.

On registered server DIMMs (RDIMM), you can have as many
as four ranks. The register chips provide buffering and fanout
for driving larger arrays, at the price of an extra clock
cycle of latency in the address/control path. Typically a
registered DIMM will use x4 chips, as a means of achieving
a higher capacity DIMM. So that's where the actual market
for x4 chips is, rather than the not-so-compatible 1GB
DDR400 high density DIMM.

Inside each memory chip, they do some similar things. They
have "banks" in there, or rectangles of memory bits. And
the third dimension inside the chip is a "bank". That's why
"rank" is used when describing the array at the memory
chip level on the DIMM. Since the name "bank" was
already taken :-)

*******

What's funny about the DIMM you just bought (16 chip x4 width),
is no company wants to document one of those :-) If I buy
a DIMM from Kingston, I can get a datasheet, where I
can see I'm getting two ranks, of 8 chips of 8x width.

But the generic companies who make the unbranded UDIMM
with the x4 chips on it, they *never* seek to document what
they've done. Instead, they couch their product
description in terms of "works with VIA, SIS, NVidia, and
not so much with Intel". Leaving the customer to
figure out how they've been snookered.

Paul

Thank you

I am very happy all the specific chipsets that work and do not work came
were documented...however since most of my older machines are Intel, I
am not going to purchase this kind of RAM again.


After I posted I Googled for hi-density vs low density
 
On 11/30/2013 4:59 PM, Paul wrote:
snip...
What's funny about the DIMM you just bought (16 chip x4 width),
is no company wants to document one of those :-) If I buy
a DIMM from Kingston, I can get a datasheet, where I
can see I'm getting two ranks, of 8 chips of 8x width.

But the generic companies who make the unbranded UDIMM
with the x4 chips on it, they *never* seek to document what
they've done. Instead, they couch their product
description in terms of "works with VIA, SIS, NVidia, and
not so much with Intel". Leaving the customer to
figure out how they've been snookered.

Paul

Do you really think that the customer has been "snookered" when the seller
says that "it will work with these but it won't work with those" and they
are being truthful? If they said nothing at all or stated that the product
worked universally then I'd question the honesty of the transaction but
when they tell the literal truth...?
 
John said:
On 11/30/2013 4:59 PM, Paul wrote:
snip...

Do you really think that the customer has been "snookered" when the
seller says that "it will work with these but it won't work with those"
and they are being truthful? If they said nothing at all or stated that
the product worked universally then I'd question the honesty of the
transaction but when they tell the literal truth...?

It's not the original buyer I'm worried about. It's
the resale market (Ebay), where the original seller
won't remember to post the disclaimer (having thrown
all docs away the day of the original purchase). Then
someone buys that crap, and they post here asking why
it doesn't work.

This is why we have so-called standards for memory.
What should be released into the channel, is stuff
that works without compromise. Then later, if some
idiot throws away their documentation, sells the
RAM on, we can be fairly certain the new purchaser
won't get snookered.

What I can't understand, is how those x4 chips are
cheaper to buy. The package they come in, I don't
think it has a lower pin count than the x8 chips. They
use a common outline for the IC package. If they charged
the same for the x4 as for the x8, then there would
be no incentive to releasing that kind of memory UDIMM
with the x4 chips on it.

*******

Mushkin used to keep a web site with some technical memos
on it. One of their testing efforts, was testing that
x4 stuff. In actual fact, for some of the so-called
compatible chipsets, you might be able to drive one stick
of the x4 stuff. But if you fill all the slots in the
motherboard with it, then it throws errors. The Mushkin
page with this information, disappeared long ago, and
Mushkin set up their site so that archive.org could not
archive the page in question. And I don't even have a
print of the test results to present as evidence. It
could be, that just the SIS chipset could drive all slots
populated that way.

*******

Price seems to over rule common sense, even with the
best of manufacturers. Kingston offers data sheets
for their RAM. Every product with a certain SKU,
should match the architecture info in the datasheet.
Yet, for one of the relatively recent products, they
broke with tradition, played the spot market for
memory chips, bought two different chip densities.
Now, when you used the Kingston search engine, it
might say that product works with my VIA motherboard,
when in fact of the two compositions they were actually
making, one of them was wrong. And all because some
days, one of those chip types was a few cents cheaper,
then the next day the other type would be cheaper.

Normally, they would make a different SKU for each
composition, and keep everyone happy. I guess someone
worked out the cost of documentation and stock keeping,
and decided it would be fun to throw out a few curve
balls to the buying public. Both compositions offer
the same capacity (1GB), but one composition used
a higher density chip than the other (and the
affected chipsets don't have the extra address
bit needed to work with the denser chip).

Paul
 
It's not the original buyer I'm worried about. It's
the resale market (Ebay), where the original seller
won't remember to post the disclaimer (having thrown
all docs away the day of the original purchase). Then
someone buys that crap, and they post here asking why
it doesn't work.

This is why we have so-called standards for memory.
What should be released into the channel, is stuff
that works without compromise. Then later, if some
idiot throws away their documentation, sells the
RAM on, we can be fairly certain the new purchaser
won't get snookered.

I'd like to know what standards the makers of retail memory
modules follow when they take 1333MHz RAM chips and use them for
modules they rate for anywhere from 1600MHz - 2666MHz (proved
by APHnetworks.com and XbitLabs.com reviews), especially when a
lot of those modules fail user testing.
 
I'd like to know what standards the makers of retail memory
modules follow when they take 1333MHz RAM chips and use them for
modules they rate for anywhere from 1600MHz - 2666MHz (proved
by APHnetworks.com and XbitLabs.com reviews), especially when a
lot of those modules fail user testing.

Who says "close enough" only works in horse shoes.
It works in the RAM business too :-)

Paul
 
Paul said:
Price seems to over rule common sense, even with the
best of manufacturers. Kingston offers data sheets
for their RAM. Every product with a certain SKU,
should match the architecture info in the datasheet.
Yet, for one of the relatively recent products, they
broke with tradition, played the spot market for
memory chips, bought two different chip densities.

Is that really anything new with Kingston? I vaguely remember having two
pairs of Kingston DIMMs in my computer years ago. Wondered how come the
modules had the same part number but different chips manufacturer,
different density and PCB... I figured it's just Kingston being
Kingston. Not a problem then or now though. USUALLY Kingston tends to be
the cheapest DIMM manufacturer in these parts so I mostly buy them.
 
Back
Top