My question is the subject of this message. Is DDR-2 the same thing as
Dual-Channel DDR? If not, what are the differences and which one should I go
with? Thank You.

No that's totally different.

Dual Channel DDR doesnt' really exist. In fact, it means that the
northbridge chipset can hold a double size bus, so before we have
64bits(single channel) then 128bits(2*64) with a dual channel.

DDR-2 is just the second generation of DDR with an enhanced design to
support higher frequencies. DDR-2 is supposed to be backward compatible with
DDR. DDR-2 start from 3200 MB/s whereas DDR range from 2100MB/s to 3200MB/s.
Therefore you could have PC3200 with DDR or DDR-2. You must check this info
before buying.


"Bradley" <(E-Mail Removed)> a écrit dans le message de news:
_pd_b.32094$(E-Mail Removed)...
> My question is the subject of this message. Is DDR-2 the same thing as
> Dual-Channel DDR? If not, what are the differences and which one should I
go
> with? Thank You.
>
>

On Mon, 23 Feb 2004 02:18:34 GMT, "Bradley" <(E-Mail Removed)> wrote:
>My question is the subject of this message. Is DDR-2 the same thing as
>Dual-Channel DDR? If not, what are the differences and which one should I go
>with? Thank You.

In a word, no. They are not the same thing and they are definitely
NOT mutually exclusive. Dual-channel DDR means, quite simply, that
there are two memory channels working together. In a more traditional
single-channel setup you install a single memory module and the system
accesses only that module. In a dual-channel setup you need to
install memory modules in pairs and the system spreads memory access
across both channels. This doubles the rate at which your processor
can shuffle data in and out of memory.

Either DDR or DDR-2 (or even the older SDRAM or RDRAM) can operate in
dual-channel systems.

DDR-2 is just the next generation of DRAM interface. It can easily
work in single or dual channel setups (or any number of other setups
used in some high-end servers). It's really only a relatively minor
update to the DDR specification that will allow for higher speeds.
While current DDR memory is having trouble getting beyond 200MHz
(400MT/s, ie the stuff sold as DDR400 or PC3200 memory), DDR-2 should
easily clock up to 333MHz (666MT/s) and possibly beyond.

At this moment, the choice between DDR and DDR-2 is pretty clear. DDR
is here, DDR-2 is not, so go with DDR. In about 2 months time the
first DDR-2 systems will begin to appear, but they will probably
command a price premium in the beginning with little to no performance
advantage (the first batch of DDR-2 memory will probably only run at
the same 200MHz as existing DDR memory, or maybe up to 266MHz at the
high-end). Eventually though (6-8 months?) DDR-2 should reach
price-parity with DDR and offer higher performance.

-------------
Tony Hill
hilla <underscore> 20 <at> yahoo <dot> ca

On Mon, 23 Feb 2004 08:55:10 +0100, "Papa Schultz" <(E-Mail Removed)> wrote:


>
>"Bradley" <(E-Mail Removed)> a écrit dans le message de news:
>_pd_b.32094$(E-Mail Removed)...
>> My question is the subject of this message. Is DDR-2 the same thing as
>> Dual-Channel DDR? If not, what are the differences and which one should I
>go
>> with? Thank You.
>>
>>
>
>No that's totally different.
>
>Dual Channel DDR doesnt' really exist. In fact, it means that the
>northbridge chipset can hold a double size bus, so before we have
>64bits(single channel) then 128bits(2*64) with a dual channel.
>
>DDR-2 is just the second generation of DDR with an enhanced design to
>support higher frequencies. DDR-2 is supposed to be backward compatible with
>DDR. DDR-2 start from 3200 MB/s whereas DDR range from 2100MB/s to 3200MB/s.
>Therefore you could have PC3200 with DDR or DDR-2. You must check this info
>before buying.

DDR-2 was not "supposed to be backward compatible with DDR".

And it most assuredly is not...

/daytripper

We dont have the same information, I read that was compatible from JETEC!
Where did you read it was not?

> DDR-2 was not "supposed to be backward compatible with DDR".
>
> And it most assuredly is not...
>
> /daytripper

On Mon, 23 Feb 2004 18:15:19 +0100, "Papa Schultz" <(E-Mail Removed)> wrote:
>
>> DDR-2 was not "supposed to be backward compatible with DDR".
>>
>> And it most assuredly is not...
>>
>> /daytripper

>We dont have the same information, I read that was compatible from JETEC!
>Where did you read it was not?

Ok, so this isn't a total waste of time, the short list of differences
includes:

DDR runs on 2.5v, uses SSTL-2.5 signaling, doesn't provide dynamic On-Die
Termination, has a smaller minimum burst length, is limited to single-ended
strobes, has "longer" pages, doesn't support Off Chip Driver calibration.

DDR2 runs on 1.8v, uses SSTL-1.8 signaling, provides dynamic ODT, has a larger
maximum burst length, supports differential strobes, has "shorter" pages,
provides OCD calibration, and has modified MRs and EMRs to support the added
features (those listed above plus a handful of diagnostic bits).

Back to your question. There may have been someone somewhere who honestly
thought he could plug DDR2 dimms into DDR platforms and operate them in
quasi-DDR mode. It begs questions about signal integrity - and certainly PLL
accuracy - at two operating voltages, never mind the economic wisdom in
support of the attempt.

But at the end of the day, surely the bottom line is: "Can you put a DDR2 dimm
into a DDR socket?"

Well, ask "JETEC" how you're going to fit the industry standard 240 pin DDR2
dimm into the industry standard 184 pin DDR socket...

/daytripper (Maybe they'll email you a really big hammer ;-)

daytripper <(E-Mail Removed)> wrote:

>DDR2 runs on 1.8v, uses SSTL-1.8 signaling, provides dynamic ODT,

How does the On-Die Termination work? Does the BIOS tell the chips
that they are or are not the last in the chain, so they can terminate
the bus or not (and is each side of a double-sided DIMM directed
differently, if one of them is at the end of the chain)?

>has a larger
>maximum burst length, supports differential strobes, has "shorter" pages,
>provides OCD calibration, and has modified MRs and EMRs to support the added
>features (those listed above plus a handful of diagnostic bits).

What's Off-Chip Driver calibration?

On Tue, 24 Feb 2004 07:45:33 -0600, chrisv <(E-Mail Removed)> wrote:

>daytripper <(E-Mail Removed)> wrote:
>
>>DDR2 runs on 1.8v, uses SSTL-1.8 signaling, provides dynamic ODT,
>
>How does the On-Die Termination work? Does the BIOS tell the chips
>that they are or are not the last in the chain, so they can terminate
>the bus or not (and is each side of a double-sided DIMM directed
>differently, if one of them is at the end of the chain)?

Much cooler than that - the key word is *dynamic*.

DDR2 provides an ODT signal for each rank of SDRAMs. During data transfer
periods (ie in real time) the On-Die Terminators for the populated ranks are
manipulated to maximize the data-phase "eye size" for significantly improved
reliability (and as these bus speeds increase, ODT becomes more and more a
"require feature").

The nature of the ODT signals are similar to CS signals and are provided by
the host memory controller (quite simple, really, as they share the exact same
state-machine that otherwise controls the ranks).

>>has a larger
>>maximum burst length, supports differential strobes, has "shorter" pages,
>>provides OCD calibration, and has modified MRs and EMRs to support the added
>>features (those listed above plus a handful of diagnostic bits).
>
>What's Off-Chip Driver calibration?

Another cool tuning "feature" that'll probably be *required* to survive decent
sized memory configurations.

OCD calibration allows the system to dynamically reduce mismatches between
output driver pull-up and pull-down strength characteristics and "tune" the
output drivers to 18 ohms to optimize signal integrity.

Using Load Mode commands to the EMR to enter an OCD drive-HIGH or drive-LOW
state, the host controller samples the voltage levels on the received "data"
and compares them to a known reference, calculates the DRAM output driver
strengths against the 18 ohm target, and then uses OCD Adjust commands to
adjust the DRAM output driver strength to achieve that goal.

Et voila, tunable transmission line drivers. Very nice.

/daytripper

daytripper <(E-Mail Removed)> wrote:

>DDR2 provides an ODT signal for each rank of SDRAMs. During data transfer
>periods (ie in real time) the On-Die Terminators for the populated ranks are
>manipulated to maximize the data-phase "eye size" for significantly improved
>reliability (and as these bus speeds increase, ODT becomes more and more a
>"require feature").

Wow. Getting to be just like RF.

In article <(E-Mail Removed)>,
(E-Mail Removed)lid says...
> daytripper <(E-Mail Removed)> wrote:
>
> >DDR2 provides an ODT signal for each rank of SDRAMs. During data transfer
> >periods (ie in real time) the On-Die Terminators for the populated ranks are
> >manipulated to maximize the data-phase "eye size" for significantly improved
> >reliability (and as these bus speeds increase, ODT becomes more and more a
> >"require feature").
>
> Wow. Getting to be just like RF.

Well, yeah. 500MHz is getting just like RF. ;-)

--
Keith