WannaKatana said:
Crucial says the correct memory is 2.2v but Gigabyte says 1.8v. Do I
take Crucial's word for it? Does this have something to do with the
issue I've read about where one has to initially boot the memory with
1.8v memory, then increase the voltage in the bios, then install the
2.2v memory?
Thanks,
Joel
If you download a datasheet for a memory chip, you can get some specs.
This spec is 7.8MB.
http://download.micron.com/pdf/datasheets/dram/ddr2/512MbDDR2.pdf
The spec for regular operation is
"VDD = +1.8V ±0.1V, VDDQ = +1.8V ±0.1V"
That is intended to give some room for motherboard regulator supply
variation. But some motherboard makers will choose to interpret that to
mean that they can supply 1.9V by default.
A second spec in the datasheet, is "Absolute Max". You should not
exceed that value, as long term, there could be damage to the
device. That Micron memory chip has an Absolute Max of 2.3V .
Not all products will necessarily list the same value.
And yes, the voltage issue is a problem. If the motherboard
makers were playing this jolly game like all the rest, then
all the motherboards would be running at 2.1V (i.e. some higher
voltage). But they don't seem to be, as otherwise there would
be fewer problems getting motherboards started.
The SPD on the DIMM, can contain a lower spec value than what is
printed on the label. That helps the user get the board started
by the "Auto" setting the first time. Then the user can bump the voltage,
change freq and timing, and raise the memory to the spec point
printed on the paper label. But the binning on a lot of the memory,
is really too tight, and there is no margin to speak of. This is
not right, because any hardware should be designed to work over
a range of temperatures, and if the memory is binned too close,
then if the room heats up, the memory may start to throw errors.
In one review I was reading on Newegg, a user was using 2.3V to
make his 2.1V memory meet its stated 4-4-4 spec. Rather than
torture the stick, he should just have sent it back and got one
that meets the spec.
This reliance on non-standard voltage, is not something every
customer is going to understand. Which is why you can see
some customer reviews, where obviously the customer doesn't
have a clue what the procedure is, to get the memory to run
error free. And it certainly isn't explained in an end
user friendly way, in the motherboard user manual.
If all the memory was binned at 1.9V, and the speed and timing
was stored in the SPD (same as the printed label), then naive
customers would not get burned. And enthusiasts could clock
to their heart's content. The industry would have a lot less
DDR2-800 to offer (since so much of it is "made" by cranking
the voltage), but at least the game would be a little more
honest, and played by JEDEC rules. But as long as the marketing
department is running the memory test machine, we'll be playing
this silly game. Standards were invented for a reason, to make
the end user experience as smooth as possible, so not
everyone has to be a "rocket scientist".
Imagine if cars had ignition timing, fuel mixture, and other tweaks
as knobs on the dash. No one would get out of their driveway in the
morning
Paul
