From: Brian V on 30 Jan 2010 17:54
PcWizard tells me so much information I need to know.
The slots can take up to 3072Mb sizes. Yes, I understand about 32-bit OS's.
I read maybe on crucial that unbuffered EEC/non-EEC is ok for my mobo.
Memory controller information says Error detection method: 64-bit EEC. The
individual RAM say EEC: No.
I would guess that the RAM pieces do not have EEC, but I can get pieces that
do have EEC?
Also read and recommened: Get RAM that is all either EEC or all non-EEC. Do
not mix. Is it impossible to mix them? Or does it depend on the motherboard?
From: Paul on 31 Jan 2010 03:53
Brian V wrote:
> PcWizard tells me so much information I need to know.
> The slots can take up to 3072Mb sizes. Yes, I understand about 32-bit OS's.
> I read maybe on crucial that unbuffered EEC/non-EEC is ok for my mobo.
> Memory controller information says Error detection method: 64-bit EEC. The
> individual RAM say EEC: No.
> I would guess that the RAM pieces do not have EEC, but I can get pieces that
> do have EEC?
> Also read and recommened: Get RAM that is all either EEC or all non-EEC. Do
> not mix. Is it impossible to mix them? Or does it depend on the motherboard?
Normally, a DIMM would house an array of memory chips forming a 64 bit wide
array. On an ECC DIMM, the array of memory chips is made 8 bits wider, for a
total of 72 bits. The extra 8 bits, holds a checksum of sorts. If you sum all
the other 8 bytes, and the total doesn't agree with the value stored in the
ninth byte, then you know a memory error has happened. The extra memory on
the module then, is adding the ability to detect errors while your data is
stored inside the RAM. (Note - the checksum thing is an analogy, and it
doesn't work exactly like that. But it uses that kind of concept. It is
a means of verification that the RAM contents haven't changed.)
The computer must support the 72 bit connection to the memory module, rather
than the regular 64 bit connection. A computer with a recent generation
of AMD Athlon64 processor, has the necessary 72 bit connection to support
ECC. But the BIOS and the OS must also provide support for the feature for
it to work. For example, if an error happens, while you're using ECC memory,
and the OS doesn't tell you about it, then that would be a waste. Or if
the BIOS hasn't turned on the ECC hardware, then that ninth extra chip
wouldn't get used.
On some motherboards, when you install ECC memory, an extra line of text
appears in one of the BIOS setup screens. It might say "ECC enable or disable?",
indicating you now have the option of turning it off if you want. If no ECC
memory is installed in the computer, you don't even get to see that
line of text, so you don't even know if the BIOS supports the operation
of the ECC feature.
In that sense, ECC is a bit annoying, in that on a desktop system, it
doesn't get very good support from the involved parties. You may know,
for example, that your hardware *should* have support, only to discover
after buying the ECC memory, that nothing is happening. Maybe you can't
see any evidence that you've installed ECC memory.
So, it is one thing to know you need the product, as a means to
enhancing the ability to detect or even correct memory errors. But
it is quite another, to satisfy all the dependencies in BIOS, hardware,
and software, to have a fully working ECC system. This might be a reason
a lot of people wouldn't bother wasting their time, since the hardware
companies aren't really helping them in this endeavor.
If using ECC memory, all the installed sticks should be ECC. That is
normally what is stated as a requirement, before you can switch on
ECC protection in the BIOS screen. In fact, the computer might beep
and show a black video screen, if you presented a mixture of sticks,
some ECC and some not. So if your current sticks are non-ECC, you'd
have to buy all new, ECC sticks, to replace them.
An AMD processor based platform, is the right place for this experiment.
Intel does make a few chipsets with ECC support, but it is damn hard
to know in advance, whether it will work or not. For example, you
might not see any messages on the BIOS screen, and then you would not
know whether this was a BIOS bug or not. AMD systems seem to be
better at supporting the feature. But on a retail computer like
your Acer, who knows what features your BIOS supports. The
sparse screen provides little in the way of details, as to
what is underneath.
This might be something that an experienced system builder
could help with, by doing testing in advance, and determining
which motherboards would be good candidates for ECC memory. It
would be more risky for an end user, to waste their money on
ECC memory, and then discover it has added nothing to their
experience. You *can* install ECC memory, as if the extra
chip is not supported by the computer, the logic signals
"flap in the breeze" and they don't hurt anything. But I
can't guarantee what every BIOS will do, when it sees the
word "ECC" in the DIMM SPD table. The DIMM has a flash
memory on it, which holds details about the memory DIMM,
and that is how the BIOS finds out what kind of DIMM has
been installed. If the BIOS sees something it doesn't like,
then it may decide to beep like crazy and not start up.
So using non-ECC memory, means a less traumatic upgrade
experience for an end user. But it also means you won't
be protected against memory errors or the detection of such.
From: Brian V on 31 Jan 2010 20:56